1 2007-02-13 Stuart Hastings <stuart@apple.com>
3 * gcc/config/i386/i386.md (fixuns_truncdfhi2): Require SSE2.
5 2007-02-13 Richard Henderson <rth@redhat.com>
7 * config/alpha/alpha.c (alpha_stdarg_optimize_hook): Strip
8 handled_component_p before looking for the indirect_ref.
10 2007-02-13 Richard Henderson <rth@redhat.com>
12 * config/i386/i386.md (bswapsi_1): Rename from bswapsi2,
14 (bswapsi2): New expander, emit code for !TARGET_BSWAP.
15 (bswaphi_lowpart): New.
16 (bswapdi2): Rename from bswapdi2_rex, remove flags clobber,
17 remove TARGET_BSWAP test. Delete expander of the same name.
19 * optabs.c (widen_bswap, expand_doubleword_bswap): New.
20 (expand_unop): Use them.
22 2007-02-13 Uros Bizjak <ubizjak@gmail.com>
24 * config/i386/i386.md (cmpdi_ccno_1_rex64, *cmpsi_ccno_1,
25 *cmphi_ccno_1, *cmpqi_ccno_1, *movsi_xor, *movstricthi_xor,
26 *movstrictqi_xor, *movdi_xor_rex64, *ashldi3_1_rex64,
27 *ashldi3_cmp_rex64, *ashldi3_cconly_rex64, ashlsi3, *ashlsi3_1_zext,
28 *ashlsi3_cmp, *ashlsi3_cconly, *ashlsi3_cmp_zext, *ashlhi3_1_lea,
29 *ashlhi3_1, *ashlhi3_cmp, *ashlhi3_cconly, *ashlqi3_1_lea,
30 *ashlqi3_1, *ashlqi3_cmp, *ashlqi3_cconly): Remove equivalent
31 assembler dialect choice from asm templates.
33 2007-02-12 Richard Henderson <rth@redhat.com>
35 * config/i386/i386.md (fixuns_trunc<SSEMODEF>si_1): New insn.
36 (fixuns_trunc<SSEMODEF>si2): Use it.
37 * config/i386/sse.md (vec_setv4sf_0): Export.
38 * config/i386/i386.c (ix86_build_const_vector): Export.
39 (ix86_split_convert_uns_si_sse): Rename from
40 ix86_expand_convert_uns_si_sse and rewrite as a splitter.
41 * config/i386/i386-protos.h: Update.
43 2007-02-13 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
46 * c-typeck.c (c_finish_return): Improve warning message.
48 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
50 * alias.c (find_symbolic_term): Delete unused function.
52 2007-02-12 Uros Bizjak <ubizjak@gmail.com>
54 * config/i386/i386.md (paritydi2, paritysi2): New expanders.
55 (paritydi2_cmp, paritydi2_cmp): New insn and split patterns.
56 (*parityhi2_cmp, *parityqi2_cmp): New insn patterns.
58 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
60 * tree.h (DECL_IGNORED_P): Document further effect for FUNCTION_DECL.
61 * cgraphunit.c (cgraph_expand_function): If DECL_IGNORED_P is set on
62 the function, temporarily point the debug interface to the null one.
64 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
66 * dwarf2out.c (round_up_to_align): New static function.
67 (field_byte_offset): Use it to round the offset.
69 2007-02-12 Richard Henderson <rth@redhat.com>
71 * config/alpha/alpha.md (bswapsi2, bswapdi2): New.
72 (inswl_const): Export.
74 2007-02-12 Richard Henderson <rth@redhat.com>
76 * calls.c (emit_library_call_value_1): If PROMOTE_MODE modifed the
77 result mode of the libcall, convert back to outmode.
79 2007-02-12 Roger Sayle <roger@eyesopen.com>
81 * config/i386/i386.md (*bswapdi2_rex): Renamed from bswapdi2.
82 (bswapdi2): New define_expand to implement 32-bit implementation.
84 2007-02-12 Nick Clifton <nickc@redhat.com>
86 * doc/invoke.texi (Overall Options): Document --help=.
87 * gcc.c (target_help_flag): Rename to print_subprocess_flag.
88 (cc1_options): Pass --help= on to cc1.
89 (display_help): Add description of --help=.
90 (process_command): Add code to handle --help=. Allow translated
91 --help and --target-help switches to be passed on to compiler
93 (main): Remove unused if statement.
94 * opts.c (columns): Remove.
95 (LEFT_COLUMN): Define.
96 (wrap_help): Add columns argument.
97 (print_filtered_help): Change parameters to be an include bitmask,
98 an exclude bitmask, an any bitmask and the column width. Move the
99 code to display the params list here. Add code to display the
100 status of options rather than their descriptions if the quiet flag
102 (print_specific_help): Change parameters to be an include bitmask,
103 an exclude bitmask and an any bitmask. Move code to look up the
104 column width here. Decide upon the title for an options listing.
105 (common_handle_options): Add code to handle --help=. Adapt code
106 for --help and --target-help to use the revised form of the
107 print_specific_help function.
108 (print_help): Delete.
109 (print_param_help): Delete.
110 (print_switch): Delete.
111 * opts.h (cl_lang_count): Add prototype.
112 (CL_PARAMS, CL_WARNING, CL_OPTIMIZATION, CL_MIN_OPTION_CLASS,
113 CL_MAX_OPTION_CLASS): New defines.
114 * optc-gen.awk: Add construction of cl_lang_count.
115 * c.opt: Add Warning attribute to warning options and Optimization
116 attribute to optimization options.
117 * common.opt: Likewise.
119 Add -fhelp and -ftarget-help as aliases for the transformed --help
120 and --target-help options.
121 * opt-functions.awk: Add code to handle Warning and Optimization
124 2007-02-12 Richard Henderson <rth@redhat.com>
126 * config/alpha/constraints.md: New file.
127 * config/alpha/alpha.c: Include tm-constrs.h.
128 (alpha_const_ok_for_letter_p, alpha_const_double_ok_for_letter_p,
129 alpha_extra_constraint): Remove.
130 (alpha_emit_conditional_branch): Use satisfies_constraint_*.
131 * config/alpha/alpha-protos.h: Update.
132 * config/alpha/alpha.h (REG_CLASS_FROM_LETTER): Remove.
133 (CONST_OK_FOR_LETTER_P, CONST_DOUBLE_OK_FOR_LETTER_P): Remove.
134 (EXTRA_CONSTRAINT): Remove.
135 * config/alpha/alpha.md: Include constraints.md.
136 (adddi splitter): Use satisfies_constraint_*.
137 * config/alpha/predicates.md (add_operand): Likewise.
138 (sext_add_operand, addition_operation): Likewise.
140 2007-02-12 Dorit Nuzman <dorit@il.ibm.com>
142 PR tree-optimization/29145
143 * tree-data-ref.c (base_addr_differ_p): Make us more conservative
144 in our handling of restrict qualified pointers.
146 2007-02-12 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
149 * doc/invoke.texi (Wunused-value): Update description.
150 (Wextra): Delete item.
151 * opts.c (set_Wextra): Don't use the value of Wextra to set the
152 value of Wunused-value.
153 * c-typeck.c (c_process_expr_stmt): Don't check extra_warnings.
154 (c_finish_stmt_expr): Don't check extra_warnings.
155 (emit_side_effect_warnings): The caller is responsible to check
158 2007-02-11 Roger Sayle <roger@eyesopen.com>
159 Matt Thomas <matt@3am-software.com>
161 * simplify-rtx.c (simplify_relational_operation_1): Correct typo.
163 2007-02-11 Roger Sayle <roger@eyesopen.com>
165 * simplify-rtx.c (simplify_relational_operation_1): Optimize
166 comparisons of POPCOUNT against zero.
167 (simplify_const_relational_operation): Likewise.
169 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
171 * doc/invoke.texi (Wextra): Delete outdated paragraph.
173 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
175 * dwarf2out.c (root_type): Delete unused function.
177 2007-02-11 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
179 * genattrtab.c (contained_in_p): Delete unused function.
180 (write_expr_attr_cache): Likewise.
182 2007-02-11 Jan Hubicka <jh@suse.cz>
184 * ipa-inline.c (cgraph_edge_badness): Add "else" missing in the
187 2007-02-11 Steven Bosscher <steven@gcc.gnu.org>
189 * fwprop.c (try_fwprop_subst): Use set_unique_reg_note
190 to add the REG_EQ* note.
191 * see.c (see_merge_one_use_extension): Likewise.
192 * local-alloc.c (update_equiv_regs): Likewise. Also don't
193 turn REG_EQUAL notes into REG_EQUIV notes if the target
194 register may have more than one set.
195 * function.c (assign_parm_setup_reg): Use set_unique_reg_note.
196 * gcse.c (try_replace_reg): Likewise.
197 * alias.c (init_alias_analysis): Use find_reg_equal_equiv_note.
198 * calls.c (fixup_tail_calls): Likewise. Abort if there is
199 more than one REG_EQUIV note.
200 * reload1.c (gen_reload): Use set_unique_reg_note.
202 2007-02-11 Uros Bizjak <ubizjak@gmail.com>
204 * config/i386/i386.c (TARGET_VECTORIZE_BUILTIN_CONVERSION): Define.
205 (ix86_builtin_conversion): New function.
207 2007-02-06 Mark Mitchell <mark@codesourcery.com>
210 * tree.h (DECL_REPLACEABLE_P): New macro.
211 * except.c (set_nothrow_function_flags): Likewise.
213 2007-02-11 Tehila Meyzels <tehila@il.ibm.com>
214 Ira Rosen <irar@il.ibm.com>
215 Dorit Nuzman <dorit@il.ibm.com>
217 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_CONVERSION): New target hook.
218 * targhooks.c (default_builtin_vectorized_conversion): New.
219 * targhooks.h (default_builtin_vectorized_function): New declaration.
220 * target.h (struct vectorize): Add builtin_conversion field.
221 * tree-vectorizer.h (type_conversion_vec_info_type): New enum
222 stmt_vec_info_type value.
223 (vectorizable_conversion): New declaration.
224 * tree-vect-analyze.c (vect_analyze_operations): Add
225 vectorizable_conversion call.
226 * target-def.h (TARGET_VECTORIZE_BUILTIN_CONVERSION): New.
227 * tree-vect-transform.c (vectorizable_conversion): New function.
228 (vect_transform_stmt): Add case for type_conversion_vec_info_type.
229 * tree-vect-generic.c (expand_vector_operations_1): Consider correct
231 * config/rs6000/rs6000.c (rs6000_builtin_conversion): New.
232 (TARGET_VECTORIZE_BUILTIN_CONVERSION): Defined.
233 (rs6000_expand_builtin): Add handling a case of ALTIVEC_BUILTIN_VCFUX
234 or ALTIVEC_BUILTIN_VCFSX.
236 2007-02-10 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
239 * pa.md (movdf): For 64-bit target, fail if operand 1 is a non-zero
240 CONST_DOUBLE and operand 0 is a hard register.
241 (movdi): For 64-bit target, remove code to force CONST_DOUBLE to
242 memory. Fail if operand 1 is a non-zero CONST_INT and operand 0
243 is a hard floating-point register.
245 2007-02-10 Richard Henderson <rth@redhat.com>
246 Jakub Jelinek <jakub@redhat.com>
247 Alexandre Oliva <aoliva@redhat.com>
249 * Makefile.in (libgcc-support, libgcc.mvars): Add emutls.c.
250 * builtin-types.def (BT_WORD): Make unsigned.
251 (BT_FN_VOID_PTR_WORD_WORD_PTR): New.
252 * builtins.def (BUILT_IN_EMUTLS_GET_ADDRESS): New.
253 (BUILT_IN_EMUTLS_REGISTER_COMMON): New.
254 * c-decl.c (grokdeclarator): Don't error if !have_tls.
255 * c-parser.c (c_parser_omp_threadprivate): Likewise.
256 * dwarf2out.c (loc_descriptor_from_tree_1): Don't do anything for
258 * expr.c (emutls_var_address): New.
259 (expand_expr_real_1): Expand emulated tls.
260 (expand_expr_addr_expr_1): Likewise.
261 * libgcc-std.ver: Add __emutls_get_address, __emutls_register_common.
262 * output.h (emutls_finish): Declare.
263 * toplev.c (compile_file): Call it.
264 * tree-ssa-address.c (gen_addr_rtx): Check for const-ness of the
265 address before wrapping in CONST.
266 * varasm.c (emutls_htab, emutls_object_type): New.
267 (EMUTLS_VAR_PREFIX, EMUTLS_TMPL_PREFIX): New.
268 (get_emutls_object_name, get_emutls_object_type): New.
269 (get_emutls_init_templ_addr, emutls_decl): New.
270 (emutls_common_1, emutls_finish): New.
271 (assemble_variable): When emulating tls, swap decls; generate
272 constructor for the emutls objects.
273 (do_assemble_alias): When emulating tls, swap decl and target name.
274 (default_encode_section_info): Don't add SYMBOL_FLAG_TLS_SHIFT
276 * varpool.c (decide_is_variable_needed): Look at force_output.
277 Recurse for emulated tls.
278 (cgraph_varpool_remove_unreferenced_decls): Remove checks redundant
279 with decide_is_variable_needed.
280 * emutls.c: New file.
281 * config/sparc/sol2.h (ASM_DECLARE_OBJECT_NAME): Only emit
282 tls_object for real tls.
284 2007-02-10 Kaz Kojima <kkojima@gcc.gnu.org>
286 PR rtl-optimization/29599
287 * reload1.c (eliminate_regs_in_insn): Take the destination
288 mode into account when computing the offset.
290 2007-02-09 Stuart Hastings <stuart@apple.com>
291 Richard Henderson <rth@redhat.com>
293 * gcc/config/i386/i386.h (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
294 * gcc/config/i386/darwin.h: (TARGET_KEEPS_VECTOR_ALIGNED_STACK): New.
295 * gcc/config/i386/i386.md (fixuns_trunc<mode>si2, fixuns_truncsfhi2,
296 fixuns_truncdfhi2): New.
297 (fix_truncsfdi_sse): Call ix86_expand_convert_sign_didf_sse.
298 (floatunsdidf2): Call ix86_expand_convert_uns_didf_sse.
299 (floatunssisf2): Add call to ix86_expand_convert_uns_sisf_sse.
300 (floatunssidf2): Allow nonimmediate source.
301 * gcc/config/i386/sse.md (movdi_to_sse): New.
302 (vec_concatv2di): Drop '*'.
303 * gcc/config/i386/i386-protos.h (ix86_expand_convert_uns_si_sse,
304 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
305 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse):
307 * gcc/config/i386/i386.c (ix86_expand_convert_uns_si_sse,
308 ix86_expand_convert_uns_didf_sse, ix86_expand_convert_uns_sidf_sse,
309 ix86_expand_convert_uns_sisf_sse, ix86_expand_convert_sign_didf_sse,
310 ix86_build_const_vector, ix86_expand_vector_init_one_nonzero): New.
311 (ix86_build_signbit_mask): Fix decl of v, refactor to call
312 ix86_build_const_vector.
313 (x86_emit_floatuns): Rewrite.
315 2007-02-10 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
317 * genautomata.c (longest_path_length): Delete unused function.
318 (struct state): Delete unused longest_path_length.
319 (UNDEFINED_LONGEST_PATH_LENGTH): Delete unused macro.
320 (get_free_state): Delete unused.
322 2007-02-09 Jan Hubicka <jh@suse.cz>
324 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 30.
325 * doc/invoke.texi (inline-unit-growth): Update default value.
327 * Makefile.in (passes.o, ipa-inline.o): Add dependencies.
328 * cgraphbuild.c (build_cgraph_edges): Compute frequencies.
329 (rebuild_cgraph_edges): Likewise.
330 * cgraph.c (cgraph_set_call_stmt): Add new argument frequency.
331 (dump_cgraph_node): Dump frequencies.
332 (cgraph_clone_edge): Add frequency scales.
333 (cgraph_clone_node): Add freuqnecy.
334 * cgraph.h (cgraph_edge): Add freuqnecy argument.
335 (CGRAPH_FREQ_BASE, CGRAPH_FREQ_MAX): New constants.
336 (cgraph_create_edge, cgraph_clone_edge, cgraph_clone_node): Update.
337 * tree-pass.h (TODO_rebuild_frequencies): New constant.
338 * cgraphunit.c (verify_cgraph_node): Verify frequencies.
339 (cgraph_copy_node_for_versioning): Update call of cgraph_clone_edge.
340 (save_inline_function_body): Likewise.
341 * ipa-inline.c: inluce rtl.h
342 (cgraph_clone_inlined_nods): Update call of cgraph_clone_node.
343 (cgraph_edge_badness): Use frequencies.
344 (cgraph_decide_recursive_inlining): Update clonning.
345 (cgraph_decide_inlining_of_small_function): Dump frequency.
346 * predict.c (estimate_bb_frequencies): Export.
347 * predict.h (estimate_bb_frequencies): Declare.
348 * tree-inline.c (copy_bb): Watch overflows.
349 (expand_call_inline): Update call of cgraph_create_edge.
350 (optimize_inline_calls): Use TODO flags to update frequnecies.
351 * passes.h: Include predict.h
352 (init_optimization_passes): Move profile ahead.
353 (execute_function_todo): Handle TODO_rebuild_frequencies.
355 2007-02-09 Roger Sayle <roger@eyesopen.com>
357 * config/alpha/alpha.c (emit_insxl): Force the first operand of
358 the insbl or inswl pattern into a register.
360 2007-02-09 Roger Sayle <roger@eyesopen.com>
362 * config/ia64/ia64.md (bswapdi2): New define_insn.
364 2007-02-09 Richard Henderson <rth@redhat.com>
366 * config/i386/constraints.md (Ym): New constraint.
367 * config/i386/i386.md (movsi_1): Change Y2 to Yi constraints.
368 (movdi_1_rex64): Split sse and xmm general register moves from
369 memory move alternatives. Use conditional register constraints.
370 (movsf_1, movdf_integer): Likewise.
371 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
372 (movdf_integer_rex64): New.
373 (pushsf_rex64): Fix output constraints.
374 * config/i386/sse.md (sse2_loadld): Split rm alternative, use Yi.
375 (sse2_stored): Likewise.
376 (sse2_storeq_rex64): New.
377 * config/i386/i386.c (x86_inter_unit_moves): Enable for not
379 (ix86_secondary_memory_needed): Don't bypass TARGET_INTER_UNIT_MOVES
380 for optimize_size. Remove SF/DFmode hack.
382 2007-02-09 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
384 * config/i386/driver-i386.c: Turn on -mtune=native for AMDFAM10.
387 2007-02-09 Nathan Sidwell <nathan@codesourcery.com>
388 Richard Sandiford <richard@codesourcery.com>
390 * config.gcc (m68010-*-netbsdelf*, m68k*-*-netbsdelf*)
391 (m68k*-*-openbsd*, m68k-*-linux*): Set default_cf_cpu.
392 (m68k-*-aout*, m68k-*-coff*, m68k-*-uclinux*, m68k-*-rtems*): Add
393 m68k/t-mlib to tmake_file.
394 (m68020-*-elf*, m68k-*-elf*): Likewise. Add t-m68kbare as well.
395 (m68k*-*-*): Use --with-arch to pick a default for --with-cpu.
396 (m680[012]0-*-*, m68k*-*-*): Add support for --with-arch.
397 Allow it to be cf or m68k. Set m68k_arch_family. If that
398 variable is not empty, add t-$m68k_arch_family to tmake_file.
399 Add t-mlibs to tmake_file.
400 * doc/install.texi: Document --with-arch=m68k and --with-arch=cf.
401 * config/m68k/t-cf: New file.
402 * config/m68k/t-m68k: Likewise.
403 * config/m68k/t-mlibs: Likewise.
404 * config/m68k/t-m68kbare (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
405 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
406 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
407 * config/m68k/t-m68kelf (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
408 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS, LIBGCC, INSTALL_LIBGCC):
410 * config/m68k/t-openbsd (MULTILIB_OPTIONS, LIBGCC): Delete.
411 (INSTALL_LIBGCC): Delete.
412 (M68K_MLIB_DIRNAMES, M68K_MLIB_OPTIONS): Define.
413 * config/m68k/t-rtems (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
414 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
415 (M68K_MLIB_CPU): Define.
416 * config/m68k/t-uclinux (MULTILIB_OPTIONS, MULTILIB_DIRNAMES)
417 (MULTILIB_MATCHES, MULTILIB_EXCEPTIONS): Delete.
418 (M68K_MLIB_CPU, M68K_MLIB_OPTIONS, M68K_MLIB_DIRNAMES): Define.
420 2007-02-09 Zdenek Dvorak <dvorakz@suse.cz>
421 Richard Guenther <rguenther@suse.de>
424 * fold-const.c (fold_comparison): Handle obfuscated comparisons
425 against INT_MIN/INT_MAX.
426 * tree-ssa-loop-ivcanon.c (remove_empty_loop): Print to dump
427 file if a loop is removed.
429 2007-02-09 Joseph Myers <joseph@codesourcery.com>
431 * calls.c (store_one_arg): Pass correct alignment to
432 emit_push_insn for non-BLKmode values.
433 * expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
434 unaligned stack slot via a suitably aligned slot.
436 2007-02-08 DJ Delorie <dj@redhat.com>
438 * config/m32c/m32c.c (m32c_unpend_compare): Add default to silence
440 (legal_subregs): Use unsigned char, make const.
441 (m32c_illegal_subreg_p): Use ARRAY_SIZE. Delete unused variables.
443 2007-02-08 Paul Brook <paul@codesourcery.com>
445 * config/arm/lib1funcs.asm (RETLDM): Pop directly into PC when no
446 special interworking needed.
448 2007-02-08 Harsha Jagasia <harsha.jagasia@amd.com>
450 * config/i386/xmmintrin.h: Make inclusion of emmintrin.h
451 conditional to __SSE2__.
452 (Entries below should have been added to first ChangeLog
453 entry for amdfam10 dated 2007-02-05)
454 * config/i386/emmintrin.h: Generate #error if __SSE2__ is not
456 * config/i386/pmmintrin.h: Generate #error if __SSE3__ is not
458 * config/i386/tmmintrin.h: Generate #error if __SSSE3__ is not
461 2007-02-08 DJ Delorie <dj@redhat.com>
463 * config/m32c/m32c-protos.h (m32c_illegal_subreg_p): New.
464 * config/m32c/m32c.c (legal_subregs): New.
465 (m32c_illegal_subreg_p): New.
466 * config/m32c/predicates.md (m32c_any_operand): Use it to reject
467 unsupported subregs of hard regs.
469 2007-02-08 Jan Hubicka <jh@suse.cz>
471 * tree-cfg.c (bsi_replace): Shortcut when replacing the statement with
472 the same one; always update histograms.
474 2007-02-08 Diego Novillo <dnovillo@redhat.com>
476 * passes.c (init_optimization_passes): Tidy comment.
478 2007-02-08 Roger Sayle <roger@eyesopen.com>
480 * simplify-rtx.c (simplify_unary_operation_1) <POPCOUNT>: We can
481 strip zero_extend, bswap and rotates from POCOUNT's argument.
482 <PARITY>: Likewise, we can strip not, bswap, sign_extend,
483 zero_extend and rotates from PARITY's argument.
484 <BSWAP>: A byte-swap followed by a byte-swap is an identity.
485 (simplify_const_unary_operation) <BSWAP>: Evaluate the byte-swap
486 of an integer constant at compile-time.
488 2007-02-08 Diego Novillo <dnovillo@redhat.com>
491 * tree-flow.h (struct var_ann_d): Remove field 'is_used'.
493 * tree-ssa-alias.c (compute_is_aliased): Remove. Update all
496 * tree-ssa-live.c (remove_unused_locals): Do not remove
497 TREE_ADDRESSABLE variables.
498 * tree-ssa-structalias.c (compute_points_to_sets): Tidy.
499 * tree-ssa-operands.c (add_virtual_operand): Remove argument
500 FOR_CLOBBER. Update all users.
501 If VAR has an associated alias set, add a virtual operand for
502 it if no alias is found to conflict with the memory reference.
504 2007-02-07 Jan Hubicka <jh@suse.cz>
505 Robert Kidd <rkidd@crhc.uiuc.edu>
507 * value-prof.c (visit_hist, free_hist): Return 1 instead of 0.
509 2007-02-07 Ian Lance Taylor <iant@google.com>
511 * lower-subreg.c (simple_move): Reject PARTIAL_INT modes.
513 2007-02-07 Roger Sayle <roger@eyesopen.com>
515 * config/rs6000/rs6000.md (ctz<mode>2, ffs<mode>2, popcount<mode>2,
516 parity<mode>2, smulsi3_highpart, abstf2_internal, allocate_stack,
517 tablejumpdi, movsi_to_cr_one): Remove constraints from
518 define_expand's match_operands.
520 2007-02-07 Roger Sayle <roger@eyesopen.com>
522 * global.c (compute_regsets): Move declatation of "i" inside of
523 #ifdef ELIMINABLE_REGS to avoid unused variable bootstrap failure.
525 2007-02-07 Jakub Jelinek <jakub@redhat.com>
528 * gimplify.c (gimplify_scan_omp_clauses): Remove special casing
529 of INDIRECT_REF <RESULT_DECL>.
531 * config/i386/i386.c (override_options): Set PTA_SSSE3 for core2.
533 2007-02-06 J"orn Rennecke <joern.rennecke@arc.com>
534 Kaz Kojima <kkojima@gcc.gnu.org>
537 * config/sh/sh.c (expand_cbranchdi4): Use scratch register
539 (sh_initialize_trampoline): Add parentheses to avoid a warning.
541 2007-02-06 Zdenek Dvorak <dvorakz@suse.cz>
543 * doc/loop.texi: Document possibility not to perform disambiguation
544 of loops with multiple latches.
545 * cfgloopmanip.c (alp_enum_p): Removed.
546 (add_loop): Handle subloops. Use get_loop_body_with_size.
547 (create_preheader): Do not allow ENTRY_BLOCK_PTR to be preheader.
548 * cfghooks.c (redirect_edge_and_branch_force): Set dominator for
549 the new forwarder block.
550 (make_forwarder_block): Only call new_bb_cbk if it is not NULL.
551 Handle the case latch is NULL.
552 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Avoid cfg modifications
553 when marking loop exits.
554 * ifcvt.c (if_convert): Ditto. Mark loop exits even if cfg cannot
556 * loop-init.c (loop_optimizer_init): Do not modify cfg. Call
557 disambiguate_loops_with_multiple_latches.
558 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Calculate dominators
559 before fix_loop_structure.
560 * cfgloop.c: Include pointer-set.h and output.h.
561 (canonicalize_loop_headers, HEADER_BLOCK, LATCH_EDGE,
562 update_latch_info, mfb_keep_just, mfb_keep_nonlatch): Removed.
563 (get_loop_latch_edges, find_subloop_latch_edge_by_profile,
564 find_subloop_latch_edge_by_ivs, find_subloop_latch_edge,
565 mfb_redirect_edges_in_set, form_subloop, merge_latch_edges,
566 disambiguate_multiple_latches, get_loop_body_with_size,
567 disambiguate_loops_with_multiple_latches): New functions.
568 (flow_loop_dump): Dump multiple latch edges.
569 (flow_loop_nodes_find): Handle loops with multiple latches.
570 (flow_loops_find): Ditto. Do not call canonicalize_loop_headers.
571 (glb_enum_p): Modified.
572 (get_loop_body): Use get_loop_body_with_size.
573 * cfgloop.h (LOOPS_HAVE_RECORDED_EXITS): New flag.
574 (AVOID_CFG_MODIFICATIONS): New constant.
575 (disambiguate_loops_with_multiple_latches, add_loop,
576 get_loop_body_with_size): Declare.
577 * Makefile.in (cfgloop.o): Add pointer-set.h and output.h.
579 2007-02-06 Seongbae Park <seongbae.park@gmail.com>
582 * global.c (compute_regsets): New function.
583 (global_alloc): Refactored ELIMINABLE_REGSET
584 and NO_GLOBAL_ALLOC_REGS computation out.
585 (rest_of_handle_global_alloc): Call compute_regsets()
586 for non-optimizing case.
588 2007-02-06 Richard Henderson <rth@redhat.com>
590 * config/i386/constraints.md (Y2): Rename from Y.
591 (Yi): New constraint.
592 * config/i386/i386.md (movsi_1, movdi_2, pushdf_nointeger,
593 pushdf_integer, movdf_nointeger, movdf_integer, zero_extendsidi2_32,
594 zero_extendsidi2_rex64, truncxfdf2_mixed): Change Y constraints to Y2.
595 (extendsfdf2_mixed, extendsfdf2_sse, truncdfsf_fast_mixed,
596 truncdfsf_fast_sse, truncdfsf_mixed, fix_truncdfdi_sse,
597 fix_truncdfsi_sse, floatsidf2_mixed, floatsidf2_sse,
598 floatdidf2_mixed, floatdidf2_sse, absnegdf2_mixed,
599 absnegdf2_sse, sse_setccdf, fop_df_comm_mixed, fop_df_comm_sse,
600 fop_df_1_mixed, fop_df_1_sse): Change Y constraints to x.
601 * config/i386/mmx.md (mov<MMXMODEI>_internal_rex64,
602 mov<MMXMODEI>_internal, movv2sf_internal_rex64, movv2sf_internal,
603 vec_extractv2si_1): Change Y constraints to Y2.
604 * config/i386/sse.md (vec_setv4sf_0, vec_concatv2df, vec_dupv4si,
605 vec_dupv2di, sse2_concatv2si, vec_concatv4si_1, vec_concatv2di):
606 Change Y constraints to Y2.
607 (sse2_loadld): Change Y constraints to x.
609 2007-02-06 Roger Sayle <roger@eyesopen.com>
611 * config/rs6000/rs6000.md (popcount<mode>2): Rewrite.
612 (parity<mode>2): New define_expand using rs6000_emit_parity.
613 * config/rs6000/rs6000.c (rs6000_emit_popcount,
614 rs6000_emit_parity): New functions.
615 * config/rs6000/rs6000-protos.h (rs6000_emit_popcount,
616 rs6000_emit_parity): Prototype here.
618 2007-02-06 Ian Lance Taylor <iant@google.com>
620 * lower-subreg.c (simple_move_operand): Reject CONST.
621 (resolve_clobber): Call validate_change rather than directly
622 assigning to XEXP (pat, 0).
624 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
626 * Makefile.in (tree-ssa-loop-ivopts.o): Add pointer-set.h dependency.
627 (tree-ssa-reassoc.o): Add pointer-set.h dependency.
628 (tree-cfg.o): Remove hashtab.h dependency.
630 * tree-ssa-loop-ivopts.c: Include pointer-set.h.
631 (struct ivopts_data): Change niters to pointer_map_t.
632 (struct nfe_cache_elt, nfe_hash, nfe_eq): Delete.
633 (niter_for_exit): Create pointer_map on demand. Change for
635 (tree_ssa_iv_optimize_init): Initialize data->niters to NULL.
636 (free_loop_data): Destroy data->niters if created and reset field.
637 (tree_ssa_iv_optimize_finalize): Don't delete data->niters here.
638 (tree_ssa_iv_optimize_loop): Check for presence of stale data.
640 * tree-ssa-reassoc.c: Include pointer-set.h.
641 (bb_rank): Change to long *.
642 (operand_rank): Change to pointer_map_t.
643 (find_operand_rank): Return long, -1 if not found. Declare as inline.
644 (insert_operand_rank): Accept long.
645 (operand_entry_hash, operand_entry_eq): Remove.
646 (get_rank): Return long. Adjust for changes above.
647 (init_reassoc): Change rank type to long. Adjust creation of bb_rank
649 (fini_reassoc): Delete operand_rank with pointer_map_destroy.
651 * tree-ssa-structalias.c (vi_for_tree): Change to pointer_map.
652 (struct tree_vi, tree_vi_t, tree_vi_hash, tree_vi_eq): Delete.
653 (insert_vi_for_tree): Rewrite for pointer_map API. Assert argument
655 (lookup_vi_for_tree): Rewrite for pointer_map API. Return varinfo_t
656 directly since it cannot be NULL.
657 (get_vi_for_tree): Rewrite for pointer_map API.
658 (find_what_p_points_to): Adjust for change to lookup_vi_for_tree.
659 (init_alias_vars): Create vi_for_tree as pointer_map.
660 (delete_points_to_sets): Delete vi_for_tree using pointer_map_destroy.
662 * tree-cfg.c: Don't include hashtab.h.
663 (edge_to_cases): Declare as pointer_map.
664 (struct edge_to_cases_elt, edge_to_cases_hash, edge_to_cases_eq):
666 (edge_to_cases_cleanup): Rewrite as pointer_map_traverse callback.
667 (start_recording_case_labels): Create edge_to_cases as pointer_map.
668 (end_recoding_case_labels): Cleanup edge_to_cases manually before
670 (record_switch_edge): Delete.
671 (get_cases_for_edge): Adjust for pointer_map API, inline
672 record_switch_edge (rewritten for new API), remove goto.
674 2006-02-06 Paolo Bonzini <bonzini@gnu.org>
676 * Makefile.in (tree-nested.o): Add pointer-set.h dependency.
677 * tree-nested.c: Include pointer-set.h.
678 (var_map_elt, var_map_eq, var_map_hash): Delete.
679 (struct nesting_info): Remove GTY marker. Change the two htab_t's
681 (nesting_info_bitmap_obstack): New.
682 (lookup_field_for_decl): Adjust for pointer_map API.
683 (lookup_tramp_for_decl): Adjust for pointer_map API.
684 (get_nonlocal_debug_decl): Adjust for pointer_map API.
685 (get_local_debug_decl): Adjust for pointer_map API.
686 (convert_nl_goto_reference): Adjust for pointer_map API.
687 (convert_nl_goto_receiver): Adjust for pointer_map API.
688 (create_nesting_tree): Create outside GGC space. Create bitmap on
689 the new obstack. Create field_map and var_map as pointer_maps.
690 (free_nesting_tree): Adjust for changes to create_nesting_tree.
692 (lower_nested_functions): Move root here, no need to NULL it.
693 Initialize and release the obstack.
695 2007-02-06 Paolo Bonzini <bonzini@gnu.org>
697 * tree.c (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p):
698 Remove prototypes and make them non-static.
699 (struct tree_int_map): Remove.
700 * tree.h (struct tree_int_map): Move here, turning TO into an
702 (tree_int_map_hash, tree_int_map_eq, tree_int_map_marked_p): Declare.
704 * tree.h (TREE_COMPLEXITY): Remove.
705 (struct tree_exp): Remove complexity field.
706 * tree.c (build1_stat): Don't set it.
708 2007-02-06 Dorit Nuzman <dorit@il.ibm.com>
709 Victor Kaplansky <victork@il.ibm.com>
711 * tree-vectorizer.c (vect_is_simple_use): Support induction.
712 (vect_is_simple_reduction): Support reduction with induction as
714 (vect_is_simple_iv_evolution): Fix formatting.
715 * tree-vect-analyze.c (vect_mark_stmts_to_be_vectorized): Fix
716 formatting. Don't mark induction phis for vectorization.
717 (vect_analyze_scalar_cycles): Analyze all inductions, then reductions.
718 * tree-vect-transform.c (get_initial_def_for_induction): New function.
719 (vect_get_vec_def_for_operand): Support induction.
720 (vect_get_vec_def_for_stmt_copy): Fix formatting and add check for
722 (vectorizable_reduction): Support reduction with induction as one of
724 (vectorizable_type_demotion): Use def-type of stmt argument rather
727 * tree-ssa-loop.c (gate_scev_const_prop): Return the value of
728 flag_tree_scev_cprop.
729 * common.opt (tree-scev-cprop): New flag.
731 * tree-vect-transform.c (vect_create_destination_var): Use 'kind' in
732 call to vect_get_new_vect_var.
734 2007-02-06 Ira Rosen <irar@il.ibm.com>
736 * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Check that
738 (vect_pattern_recog_1): Likewise.
740 2007-02-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
742 * fold-const.c (negate_expr_p): Handle CONJ_EXPR.
743 (fold_negate_expr): Likewise.
745 2007-02-05 Alexandre Oliva <aoliva@redhat.com>
748 * dwarf2out.c (modified_type_die): Follow DECL_ORIGINAL_TYPE
749 even if cv-qualification is the same.
751 2007-02-05 Geoffrey Keating <geoffk@apple.com>
753 * config/rs6000/darwin-tramp.asm (__trampoline_setup): Call
754 __enable_execute_stack on completion.
756 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
758 * config/i386/athlon.md (athlon_fldxf_k8, athlon_fld_k8,
759 athlon_fstxf_k8, athlon_fst_k8, athlon_fist, athlon_fmov,
760 athlon_fadd_load, athlon_fadd_load_k8, athlon_fadd, athlon_fmul,
761 athlon_fmul_load, athlon_fmul_load_k8, athlon_fsgn,
762 athlon_fdiv_load, athlon_fdiv_load_k8, athlon_fdiv_k8,
763 athlon_fpspc_load, athlon_fpspc, athlon_fcmov_load,
764 athlon_fcmov_load_k8, athlon_fcmov_k8, athlon_fcomi_load_k8,
765 athlon_fcomi, athlon_fcom_load_k8, athlon_fcom): Added amdfam10.
767 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
769 * config/i386/i386.md (x86_sahf_1, cmpfp_i_mixed, cmpfp_i_sse,
770 cmpfp_i_i387, cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387,
771 swapsi, swaphi_1, swapqi_1, swapdi_rex64, fix_truncsfdi_sse,
772 fix_truncdfdi_sse, fix_truncsfsi_sse, fix_truncdfsi_sse,
773 x86_fldcw_1, floatsisf2_mixed, floatsisf2_sse, floatdisf2_mixed,
774 floatdisf2_sse, floatsidf2_mixed, floatsidf2_sse,
775 floatdidf2_mixed, floatdidf2_sse, muldi3_1_rex64, mulsi3_1,
776 mulsi3_1_zext, mulhi3_1, mulqi3_1, umulqihi3_1, mulqihi3_insn,
777 umulditi3_insn, umulsidi3_insn, mulditi3_insn, mulsidi3_insn,
778 umuldi3_highpart_rex64, umulsi3_highpart_insn,
779 umulsi3_highpart_zext, smuldi3_highpart_rex64,
780 smulsi3_highpart_insn, smulsi3_highpart_zext, x86_64_shld,
781 x86_shld_1, x86_64_shrd, sqrtsf2_mixed, sqrtsf2_sse,
782 sqrtsf2_i387, sqrtdf2_mixed, sqrtdf2_sse, sqrtdf2_i387,
783 sqrtextendsfdf2_i387, sqrtxf2, sqrtextendsfxf2_i387,
784 sqrtextenddfxf2_i387): Added amdfam10_decode.
786 * config/i386/athlon.md (athlon_idirect_amdfam10,
787 athlon_ivector_amdfam10, athlon_idirect_load_amdfam10,
788 athlon_ivector_load_amdfam10, athlon_idirect_both_amdfam10,
789 athlon_ivector_both_amdfam10, athlon_idirect_store_amdfam10,
790 athlon_ivector_store_amdfam10): New define_insn_reservation.
791 (athlon_idirect_loadmov, athlon_idirect_movstore): Added
794 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
796 * config/i386/athlon.md (athlon_call_amdfam10,
797 athlon_pop_amdfam10, athlon_lea_amdfam10): New
798 define_insn_reservation.
799 (athlon_branch, athlon_push, athlon_leave_k8, athlon_imul_k8,
800 athlon_imul_k8_DI, athlon_imul_mem_k8, athlon_imul_mem_k8_DI,
801 athlon_idiv, athlon_idiv_mem, athlon_str): Added amdfam10.
803 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
805 * config/i386/athlon.md (athlon_sseld_amdfam10,
806 athlon_mmxld_amdfam10, athlon_ssest_amdfam10,
807 athlon_mmxssest_short_amdfam10): New define_insn_reservation.
809 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
811 * config/i386/athlon.md (athlon_sseins_amdfam10): New
812 define_insn_reservation.
813 * config/i386/i386.md (sseins): Added sseins to define_attr type
814 and define_attr unit.
815 * config/i386/sse.md: Set type attribute to sseins for insertq
818 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
820 * config/i386/athlon.md (sselog_load_amdfam10, sselog_amdfam10,
821 ssecmpvector_load_amdfam10, ssecmpvector_amdfam10,
822 ssecomi_load_amdfam10, ssecomi_amdfam10,
823 sseaddvector_load_amdfam10, sseaddvector_amdfam10): New
824 define_insn_reservation.
825 (ssecmp_load_k8, ssecmp, sseadd_load_k8, seadd): Added amdfam10.
827 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
829 * config/i386/athlon.md (cvtss2sd_load_amdfam10,
830 cvtss2sd_amdfam10, cvtps2pd_load_amdfam10, cvtps2pd_amdfam10,
831 cvtsi2sd_load_amdfam10, cvtsi2ss_load_amdfam10,
832 cvtsi2sd_amdfam10, cvtsi2ss_amdfam10, cvtsd2ss_load_amdfam10,
833 cvtsd2ss_amdfam10, cvtpd2ps_load_amdfam10, cvtpd2ps_amdfam10,
834 cvtsX2si_load_amdfam10, cvtsX2si_amdfam10): New
835 define_insn_reservation.
837 * config/i386/sse.md (cvtsi2ss, cvtsi2ssq, cvtss2si,
838 cvtss2siq, cvttss2si, cvttss2siq, cvtsi2sd, cvtsi2sdq,
839 cvtsd2si, cvtsd2siq, cvttsd2si, cvttsd2siq,
840 cvtpd2dq, cvttpd2dq, cvtsd2ss, cvtss2sd,
841 cvtpd2ps, cvtps2pd): Added amdfam10_decode attribute.
843 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
845 * config/i386/athlon.md (athlon_ssedivvector_amdfam10,
846 athlon_ssedivvector_load_amdfam10, athlon_ssemulvector_amdfam10,
847 athlon_ssemulvector_load_amdfam10): New define_insn_reservation.
848 (athlon_ssediv, athlon_ssediv_load_k8, athlon_ssemul,
849 athlon_ssemul_load_k8): Added amdfam10.
851 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
853 * config/i386/i386.h (TARGET_SSE_UNALIGNED_MOVE_OPTIMAL): New macro.
854 (x86_sse_unaligned_move_optimal): New variable.
856 * config/i386/i386.c (x86_sse_unaligned_move_optimal): Enable for
858 (ix86_expand_vector_move_misalign): Add code to generate movupd/movups
859 for unaligned vector SSE double/single precision loads for AMDFAM10.
861 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
863 * config/i386/i386.h (TARGET_AMDFAM10): New macro.
864 (TARGET_CPU_CPP_BUILTINS): Add code for amdfam10.
865 Define TARGET_CPU_DEFAULT_amdfam10.
866 (TARGET_CPU_DEFAULT_NAMES): Add amdfam10.
867 (processor_type): Add PROCESSOR_AMDFAM10.
869 * config/i386/i386.md: Add amdfam10 as a new cpu attribute to match
870 processor_type in config/i386/i386.h.
871 Enable imul peepholes for TARGET_AMDFAM10.
873 * config.gcc: Add support for --with-cpu option for amdfam10.
875 * config/i386/i386.c (amdfam10_cost): New variable.
876 (m_AMDFAM10): New macro.
877 (m_ATHLON_K8_AMDFAM10): New macro.
878 (x86_use_leave, x86_push_memory, x86_movx, x86_unroll_strlen,
879 x86_cmove, x86_3dnow_a, x86_deep_branch, x86_use_simode_fiop,
880 x86_promote_QImode, x86_integer_DFmode_moves,
881 x86_partial_reg_dependency, x86_memory_mismatch_stall,
882 x86_accumulate_outgoing_args, x86_arch_always_fancy_math_387,
883 x86_sse_partial_reg_dependency, x86_sse_typeless_stores,
884 x86_use_ffreep, x86_use_incdec, x86_four_jump_limit,
885 x86_schedule, x86_use_bt, x86_cmpxchg16b, x86_pad_returns):
886 Enable/disable for amdfam10.
887 (override_options): Add amdfam10_cost to processor_target_table.
888 Set up PROCESSOR_AMDFAM10 for amdfam10 entry in
889 processor_alias_table.
890 (ix86_issue_rate): Add PROCESSOR_AMDFAM10.
891 (ix86_adjust_cost): Add code for amdfam10.
893 2007-02-05 Harsha Jagasia <harsha.jagasia@amd.com>
895 * config/i386/i386.opt: Add new Advanced Bit Manipulation (-mabm)
896 instruction set feature flag. Add new (-mpopcnt) flag for popcnt
897 instruction. Add new SSE4A (-msse4a) instruction set feature flag.
898 * config/i386/i386.h: Add builtin definition for SSE4A.
899 * config/i386/i386.md: Add support for ABM instructions
901 * config/i386/sse.md: Add support for SSE4A instructions
902 (movntss, movntsd, extrq, insertq).
903 * config/i386/i386.c: Add support for ABM and SSE4A builtins.
904 Add -march=amdfam10 flag.
905 * config/i386/ammintrin.h: Add support for SSE4A intrinsics.
906 * doc/invoke.texi: Add documentation on flags for sse4a, abm, popcnt
908 * doc/extend.texi: Add documentation for SSE4A builtins.
910 2007-02-05 Bob Wilson <bob.wilson@acm.org>
912 * config/xtensa/xtensa.c (constantpool_mem_p): Skip over SUBREGs.
914 2007-02-05 Richard Guenther <rguenther@suse.de>
916 * tree-vectorizer.h (vectorizable_function): Add argument type
917 argument, change return type.
918 * tree-vect-patterns.c (vect_recog_pow_pattern): Adjust caller.
919 * tree-vect-transform.c (vectorizable_function): Handle extra
920 argument, return vectorized function decl.
921 (build_vectorized_function_call): Remove.
922 (vectorizable_call): Handle calls with result and argument types
923 differing. Handle loop vectorization factor correctly.
924 * targhooks.c (default_builtin_vectorized_function): Adjust for
926 * targhooks.h (default_builtin_vectorized_function): Likewise.
927 * target.h (builtin_vectorized_function): Add argument type
929 * config/i386/i386.c (ix86_builtin_vectorized_function): Handle
930 extra argument, allow vectorizing of lrintf.
931 * doc/tm.texi (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Adjust
932 documentation of target hook.
934 2007-02-05 Hans-Peter Nilsson <hp@axis.com>
937 * config/cris/cris.md ("*andsi_movu", "*andsi_clear", "*andhi_movu")
938 ("*andhi_clear", andu (casesi+45)): For size-changed operand where
939 memory is allowed, require !side_effects_p, not just !MEM_VOLATILE_P.
941 2007-02-05 Roger Sayle <roger@eyesopen.com>
943 * fold-const.c (fold_unary) <REAL_PART>: Test for availability of
944 BUILT_IN_COS before simplifying REAL_PART(CEXPI)) to COS.
945 <IMAG_PART>: Likewise, check for availability of BUILT_IN_SIN.
946 * builtins.c (fold_builtin_sincos): Check for TARGET_C99_FUNCTIONS
947 before canonicalizing sincos to cexpi.
948 (fold_builtin_cexp): Likewise, for canonicalizing cexp to cexpi.
950 2007-02-05 Roger Sayle <roger@eyesopen.com>
952 * config/alpha/alpha.c (alpha_add_builtins): New Helper function.
953 Set TREE_READONLY and TREE_NOTHROW directly, not via attributes.
954 (alpha_init_builtins): Use alpha_add_builtins to process tables.
956 2007-02-05 Roger Sayle <roger@eyesopen.com>
958 * mips-tfile.c (initialize_init_file): Correct endianness test.
960 2007-02-05 Kazu Hirata <kazu@codesourcery.com>
962 * config/m68k/m68k.md (pushdi-1, pushdi, movsi+1): Don't use
965 2007-02-05 Richard Sandiford <richard@codesourcery.com>
967 * dwarf2out.c (dwarf2out_frame_debug_expr): Record the register
968 saves in a PARALLEL before the register assignments.
970 2007-02-05 Richard Sandiford <richard@codesourcery.com>
972 * doc/tm.texi (DWARF_ALT_FRAME_RETURN_COLUMN): Do not require
973 DWARF_FRAME_RETURN_COLUMN to be a general register.
974 * dwarf2out.c (init_return_column_size): New function, split from...
975 (expand_builtin_init_dwarf_reg_sizes): ...here. Allow both
976 DWARF_FRAME_RETURN_COLUMN and DWARF_ALT_FRAME_RETURN_COLUMN
977 to be nongeneral registers.
978 * config/m68k/m68k.h (DWARF_FRAME_REGNUM): Only map FP and
980 (DWARF_FRAME_REGISTERS, DWARF_FRAME_RETURN_COLUMN): Define.
981 (DWARF_ALT_FRAME_RETURN_COLUMN): Define.
983 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
985 * cfgcleanup.c (try_optimize_cfg): Avoid removing ENTRY_BLOCK_PTR.
987 2007-02-04 Zdenek Dvorak <dvorakz@suse.cz>
989 * cfgloopmanip.c (loop_delete_branch_edge): Removed.
990 (remove_path): Use can_remove_branch_p and remove_branch instead
991 of loop_delete_branch_edge.
992 * tree-ssa-loop-manip.c (scale_dominated_blocks_in_loop): New function.
993 (tree_transform_and_unroll_loop): Remove dead branches immediately.
994 Update profile using scale_dominated_blocks_in_loop.
995 * cfghooks.c (can_remove_branch_p, remove_branch): New functions.
996 * cfghooks.h (struct cfg_hooks): Add can_remove_branch_p.
997 (can_remove_branch_p, remove_branch): Declare.
998 * tree-cfg.c (tree_can_remove_branch_p): New function.
999 (tree_cfg_hooks): Add tree_can_remove_branch_p.
1000 * cfgrtl.c (rtl_can_remove_branch_p): New function.
1001 (rtl_cfg_hooks, cfg_layout_rtl_cfg_hook): Add rtl_can_remove_branch_p.
1003 2007-02-05 Jan Hubicka <jh@suse.cz>
1006 * ipa-inline.c (cgraph_clone_inlined_nodes): When there are unanalyzed
1007 nodes in cgraph, don't remove offline copy of the function.
1009 2007-02-04 Jan Hubicka <jh@suse.cz>
1011 * tree-sra.c (sra_walk_expr): Add linebreaks. BITFIELD_REFs into
1012 vectors might cause maybe_lookup_element_for_expr to be called
1013 on non-sra-candidate.
1015 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1017 * config/bfin/bfin-modes.def, config/bfin/bfin.c,
1018 config/bfin/bfin.md, config/bfin/predicates.md: Follow
1019 spelling conventions.
1021 2007-02-04 Richard Guenther <rguenther@suse.de>
1024 * fold-const.c (try_move_mult_to_index): Make sure to not
1025 overflow one dimension of a multi-dimensional array access.
1027 2007-02-04 Jan Hubicka <jh@suse.cz>
1029 * passes.c (init_optimization_passes): Reindent.
1031 2007-02-04 Jan Hubicka <jh@suse.cz>
1032 Eric Botcazou <ebotcazou@adacore.com>
1034 * tree-optimize.c (has_abnormal_outgoing_edge_p): Move to...
1035 (execute_fixup_cfg): Break out the abnormal goto code.
1036 * tree-inline.c (has_abnormal_outgoing_edge_p): ...here.
1037 (make_nonlocal_label_edges): Move here from execute_fixup_cfg.
1038 (optimize_inline_calls): Call make_nonlocal_label_edges.
1040 2007-02-04 Jan Hubicka <jh@suse.cz>
1042 * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Return
1043 true when something was changed.
1044 (rename_ssa_copies): When something was changed, do
1045 TODO_remove_unused_locals.
1046 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1047 add TODO_remove_unused_locals when instruction was removed.
1049 2007-02-04 Jan Hubicka <jh@suse.cz>
1051 * ipa-inline.c (try_inline): Improve debug output; work on already
1053 (cgraph_decide_inlining_incrementally): Indent; improve debug output;
1054 call try_inline for already inlined edges too when flattening;
1055 inline also functions that make callee growth but overall unit size
1058 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1060 * config/m32c/bitops.md, config/m32c/jump.md,
1061 config/m32c/m32c.c, config/m32c/m32c.h, config/m32r/m32r.c,
1062 config/m32r/m32r.h, config/m32r/m32r.md,
1063 config/m32r/predicates.md, config/m68hc11/larith.asm,
1064 config/m68hc11/m68hc11.c, config/m68hc11/m68hc11.h,
1065 config/m68k/m68k.h, config/mcore/mcore.md, config/mips/4k.md,
1066 config/mips/mips-protos.h, config/mips/mips.c,
1067 config/mips/mips.h, config/mips/mips.md, config/mips/mips16.S,
1068 config/mn10300/mn10300.h, config/mn10300/predicates.md,
1069 config/mt/mt.c, config/mt/mt.h, config/mt/mt.md: Follow
1070 spelling conventions.
1072 * config/v850/v850.c, config/v850/v850.h, config/v850/v850.md:
1073 Follow spelling conventions.
1075 2007-02-03 Douglas Gregor <doug.gregor@gmail.com>
1077 * c-opts.c (c_common_post_options): If C++0x mode is enabled, don't
1078 warn about C++0x compatibility.
1080 2007-02-04 Kazu Hirata <kazu@codesourcery.com>
1082 * config/h8300/h8300.c, config/h8300/h8300.h,
1083 config/h8300/h8300.md: Follow spelling conventions.
1085 2007-02-03 Uros Bizjak <ubizjak@gmail.com>
1088 * combine.c (try_combine): Do not substitute source operand
1089 with constants wider than 2 * HOST_BITS_PER_WIDE_INT.
1091 2007-02-03 Jan Hubicka <jh@suse.cz>
1093 PR gcov-profile/30650
1094 * value-prof.c (stringop_block_profile): Fix handling of size counter;
1095 do not divide by zero for never executed counters.
1096 (tree_find_values_to_profile): Fix counters.
1097 * gcov-ui.h (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): Fix comments.
1099 2007-02-03 Ian Lance Taylor <iant@google.com>
1101 * lower-subreg.c (simple_move_operand): New static function,
1102 broken out of simple_move. Reject LABEL_REF, SYMBOL_REF, and HIGH
1104 (simple_move): Call simple_move_operand.
1105 (find_decomposable_subregs): Add special handling of MEMs.
1106 (can_decompose_p): Rename from cannot_decompose_p. Reverse
1107 meaning of return value. If we see a hard register, test whether
1108 it can store a word_mode value. Change all callers.
1110 2007-02-03 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1112 * pa.md (addvdi3, addvsi3, subvdi3, subvsi3, negvdi2, negvsi2): New
1113 ftrapv insns and expanders.
1114 (subdi3): Change define_expand operand 1 to arith11_operand, and
1115 operand 2 to reg_or_0_operand. Change constraints of 64-bit insn
1116 pattern to handle reg_or_0 operands. Revise 32-bit insn pattern to
1117 handle 11-bit constants and reg_or_0 operands in operands 1 and 2,
1121 * varasm.c (notice_global_symbol): Treat global objects as weak when
1124 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1126 * emit-rtl.c (dconstpi): Delete.
1128 (init_emit_once): Delete dconstpi and init dconstsqrt2.
1129 * real.h (dconstpi): Delete.
1131 * builtins.c (fold_builtin_cabs): Use dconstsqrt2.
1132 (fold_builtin_hypot): Likewise.
1134 2007-02-03 Tom Tromey <tromey@redhat.com>
1137 * gcc.c (cpp_unique_options): Any of -ggdb3, -gstabs3,
1138 -gcoff3, -gxcoff3, -gvms3 implies -dD.
1140 2007-02-03 Kazu Hirata <kazu@codesourcery.com>
1142 * c-decl.c, config/avr/avr.c, config/avr/avr.h,
1143 config/m68k/m68k.c, config/m68k/netbsd-elf.h,
1144 config/mn10300/mn10300.c, config/pdp11/pdp11.h,
1145 config/rs6000/cell.md, config/rs6000/darwin.h,
1146 config/sh/sh.md, config/sh/sh4-300.md, config/spu/spu.c,
1147 config/spu/spu.md, cselib.c, expr.c, haifa-sched.c, hwint.h,
1148 jump.c, reload.c, sched-deps.c, sched-int.h, tree-inline.c,
1149 tree-profile.c, tree-ssa-live.h, tree-vrp.c: Fix comment
1150 typos. Follow spelling conventions.
1151 * doc/invoke.texi: Follow spelling conventions.
1153 2007-02-03 Roger Sayle <roger@eyesopen.com>
1155 * simplify-rtx.c (simplify_relational_operation_1): Implement some
1156 canonicalization transformations that attempt to simplify integer
1157 constant comparisons to become comparisons against zero.
1159 2007-02-03 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1161 * builtins.c (fold_builtin_cabs): Fold cabs(x+xi) into
1163 * fold-const.c (fold_binary): Fix comment typos. Fold complex
1164 (x,0)-(0,y) into (x,-y). Likewise (0,y)-(x,0) into (-x,y).
1166 2007-02-02 Mike Stump <mrs@apple.com>
1168 * config/darwin9.h (DARWIN_LINKER_GENERATES_ISLANDS): Add.
1169 * config/rs6000/rs6000.c (DARWIN_GENERATE_ISLANDS): Add.
1170 (output_call): Use DARWIN_GENERATE_ISLANDS to decide when to
1171 generate a branch island.
1173 2007-02-02 Bob Wilson <bob.wilson@acm.org>
1175 * config/xtensa/xtensa.c (smalloffset_mem_p): Use BASE_REG_P.
1176 (xtensa_legitimate_address_p): New.
1177 (xtensa_legitimize_address): New.
1178 (xtensa_output_addr_const_extra): New.
1179 * config/xtensa/xtensa.h (REG_OK_STRICT_FLAG): Define.
1181 (REG_OK_FOR_BASE_P): Use BASE_REG_P.
1182 (GO_IF_LEGITIMATE_ADDRESS): Move code to xtensa_legitimate_address_p.
1183 (LEGITIMIZE_ADDRESS): Move code to xtensa_legitimize_address.
1184 (OUTPUT_ADDR_CONST_EXTRA): Move code to xtensa_output_addr_const_extra.
1185 * config/xtensa/xtensa-protos.h (xtensa_legitimate_address_p): New.
1186 (xtensa_legitimize_address): New.
1187 (xtensa_output_addr_const_extra): New.
1189 2007-02-02 Steve Ellcey <sje@cup.hp.com>
1191 * config/ia64/ia64.c (ia64_print_operand): Fix compare strings.
1193 2007-02-02 Ian Lance Taylor <iant@google.com>
1195 * expmed.c (expand_divmod): Add comment.
1197 2007-02-02 Kazu Hirata <kazu@codesourcery.com>
1199 * emit-rtl.c (renumber_insns): Remove.
1200 * flags.h: Remove the extern for flag_renumber_insns.
1201 * rtl.h: Remove the prototype for renumber_insns.
1202 * toplev.c (flag_renumber_insns): Remove.
1204 2007-02-02 Hui-May Chang <hm.chang@apple.com>
1206 Revert for x86 darwin:
1207 2005-06-19 Uros Bizjak <uros@kss-loka.si>
1209 * config/i386/i386.c (ix86_function_arg_regno_p): Put back the
1210 code before the following patch under TARGET_MACHO.
1211 (ix86_function_value_regno_p): Likewise.
1213 2007-02-02 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1215 * fold-const.c (negate_expr_p, fold_negate_expr): Handle
1218 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1221 * config/ia64/ia64.c (ia64_speculate_insn): Restrict to memory loads to
1222 general or fp registers. Add comments.
1223 * config/ia64/ia64.md (reg_pred_prefix): Add comment.
1225 2007-02-02 Paolo Bonzini <bonzini@gnu.org>
1227 * pointer-set.c (insert_aux): Only return insertion slot.
1228 (pointer_set_insert): Adjust.
1229 (pointer_set_traverse, struct pointer_map_t, pointer_map_create,
1230 pointer_map_destroy, pointer_map_insert, pointer_map_contains,
1231 pointer_map_traverse): New.
1232 * pointer-set.h (pointer_set_traverse, struct pointer_map_t,
1233 pointer_map_create, pointer_map_destroy, pointer_map_insert,
1234 pointer_map_contains, pointer_map_traverse): Declare.
1236 2007-02-02 Jakub Jelinek <jakub@redhat.com>
1239 * builtins.c (fold_builtin_sprintf): Do not attempt to optimize
1240 sprintf (str, "%s"). Do not optimize sprintf (str, "nopercent", p++).
1242 2007-02-02 Maxim Kuvyrkov <mkuvyrkov@ispras.ru>
1244 * sched-int.h (ds_to_dk, dk_to_ds): Declare functions.
1246 (struct _dep): New type.
1247 (dep_t): New typedef.
1248 (DEP_PRO, DEP_CON, DEP_KIND): New access macros.
1249 (DEP_STATUS): New access macro. The macro with the same name was
1250 renamed to DEP_LINK_STATUS.
1251 (dep_init): Declare function
1253 (struct _dep_link): New type.
1254 (dep_link_t): New typedef.
1255 (DEP_LINK_NODE, DEP_LINK_NEXT, DEP_LINK_PREV_NEXTP): New access macros.
1256 (DEP_LINK_DEP, DEP_LINK_PRO, DEP_LINK_CON, DEP_LINK_KIND): New macros.
1257 (DEP_LINK_STATUS): New macro.
1258 (debug_dep_links): New debug function.
1260 (struct _deps_list): New type.
1261 (deps_list_t): New typedef.
1262 (DEPS_LIST_FIRST): New access macro.
1263 (FOR_EACH_DEP_LINK): New cycle macro.
1264 (create_deps_list, free_deps_list, delete_deps_list): Declare
1266 (deps_list_empty_p, debug_deps_list, add_back_dep_to_deps_list): Ditto.
1267 (find_link_by_pro_in_deps_list, find_link_by_con_in_deps_list): Ditto.
1268 (copy_deps_list_change_con): Ditto.
1270 (move_dep_link): Declare function.
1272 (struct _dep_node): New type.
1273 (dep_node_t): New typedef.
1274 (DEP_NODE_BACK, DEP_NODE_DEP, DEP_NODE_FORW): New access macros.
1276 (struct haifa_insn_data.back_deps): New field to hold backward
1277 dependencies of the insn.
1278 (struct haifa_insn_data.depend): Rename to forw_deps. Change its type
1280 (struct haifa_insn_data.resolved_deps): Rename to resolved_back_deps.
1281 Change its type to deps_list_t.
1282 (INSN_BACK_DEPS): New access macro to use instead of LOG_LINKS.
1283 (INSN_DEPEND): Rename to INSN_FORW_DEPS.
1284 (RESOLVED_DEPS): Rename to INSN_RESOLVED_BACK_DEPS.
1286 (INSN_COST): Move to haifa-sched.c. Use insn_cost () instead.
1288 (DEP_STATUS): Rename to DEP_LINK_STATUS. Fix typo in the comment.
1290 (add_forw_dep, delete_back_forw_dep, insn_cost): Update declaration and
1292 (dep_cost): Declare.
1294 * sched-deps.c (CHECK): New macro to (en/dis)able sanity checks.
1295 (ds_to_dk, dk_to_ds): New functions.
1297 (init_dep_1): New static function.
1298 (init_dep): New function.
1299 (copy_dep): New static function.
1301 (dep_link_consistent_p, attach_dep_link, add_to_deps_list): New static
1303 (detach_dep_link): New static function.
1304 (move_dep_link): New function.
1306 (dep_links_consistent_p, dump_dep_links): New static functions.
1307 (debug_dep_links): New debugging function.
1309 (deps_obstack, dl_obstack, dn_obstack): New static variables.
1311 (alloc_deps_list, init_deps_list): New static functions.
1312 (create_deps_list): New function.
1313 (clear_deps_list): New static function.
1314 (free_deps_list, delete_deps_list, deps_list_empty_p): New functions.
1315 (deps_list_consistent_p, dump_deps_list): New static functions.
1316 (debug_deps_list): New function.
1317 (add_back_dep_to_deps_list, find_link_by_pro_in_deps_list): New
1319 (find_link_by_con_in_deps_list, copy_deps_list_change_con): Ditto.
1321 (maybe_add_or_update_back_dep_1, add_or_update_back_dep_1): Update to
1322 use new scheduler dependencies lists.
1323 (add_back_dep, delete_all_dependences, fixup_sched_groups): Ditto.
1324 (sched_analyze): Ditto. Initialize dependencies lists.
1325 (add_forw_dep, compute_forward_dependences): Update to use new
1326 scheduler dependencies lists.
1328 (init_dependency_caches): Init deps_obstack.
1329 (free_dependency_caches): Free deps_obstack.
1331 (adjust_add_sorted_back_dep, adjust_back_add_forw_dep): Update to use
1332 new scheduler dependencies lists.
1333 (delete_forw_dep, add_or_update_back_forw_dep): Ditto.
1334 (add_back_forw_dep, delete_back_forw_dep): Ditto.
1336 * sched-rgn.c (set_spec_fed, find_conditional_protection, is_pfree):
1337 Update to use new scheduler dependencies lists.
1338 (is_conditionally_protected, is_prisky, add_branch_dependences): Ditto.
1339 (debug_dependencies): Ditto.
1340 (schedule_region): Update comments.
1342 * sched-ebb.c (earliest_block_with_similiar_load): Update to use new
1343 scheduler dependencies lists.
1344 (schedule_ebb): Update comments.
1346 * rtl.def (DEPS_LIST): Remove.
1348 * lists.c (unused_deps_list): Remove.
1349 (free_list): Update assertions.
1351 (alloc_DEPS_LIST, free_DEPS_LIST_list, free_DEPS_LIST_node): Remove.
1352 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1354 * rtl.h (free_DEPS_LIST_list, alloc_DEPS_LIST): Remove declarations.
1355 (remove_free_DEPS_LIST_elem, copy_DEPS_LIST_list): Ditto.
1357 * haifa-sched.c (comments): Update.
1358 (insn_cost1): Remove. Inline the code into insn_cost ().
1359 (insn_cost): Update to use new scheduler dependencies lists. Move
1360 processing of the dependency cost to dep_cost ().
1361 (dep_cost): New function. Use it instead of insn_cost () when
1362 evaluating cost of the dependency. Use compatible interface to
1363 interact with the target.
1364 (priority): Update to use new scheduler dependencies lists.
1365 (rank_for_schedule): Ditto. Optimize heuristic that prefers the insn
1366 with greater number of insns that depend on the insn.
1367 (schedule_insn): Update to use new scheduler dependencies lists. Add
1368 code to free backward dependencies lists. Inline and optimize code
1369 from resolve_dep () - see PR28071.
1370 (ok_for_early_queue_removal): Update to use new scheduler dependencies
1371 lists. Update call to targetm.sched.is_costly_dependence hook.
1373 (fix_inter_tick, try_ready, fix_tick_ready): Update to use new
1374 scheduler dependencies lists.
1376 (resolve_dep): Remove. Move the logic to schedule_insn ().
1377 (init_h_i_d): Initialize dependencies lists.
1379 (process_insn_depend_be_in_spec): Rename to
1380 process_insn_forw_deps_be_in_spec. Update to use new scheduler
1382 (add_to_speculative_block, create_check_block_twin, fix_recovery_deps):
1383 Update to use new scheduler dependencies lists.
1384 (clear_priorities, calc_priorities, add_jump_dependencies): Ditto.
1386 * ddg.c (create_ddg_dependence, create_ddg_dep_no_link): Update to use
1387 new scheduler dependencies lists.
1388 (build_intra_loop_deps): Ditto.
1390 * target.h (struct _dep): Declare to use in
1391 gcc_target.sched.is_costly_dependence.
1392 (struct gcc_target.sched.adjust_cost): Fix typo.
1393 (struct gcc_target.sched.is_costly_dependence): Change signature to use
1394 single dep_t parameter instead of an equivalent triad.
1395 (struct gcc_target.sched.adjust_cost_2): Remove.
1397 * target-def.h (TARGET_SCHED_ADJUST_COST_2): Remove.
1399 * reg-notes.def (DEP_TRUE, DEP_OUTPUT, DEP_ANTI): Update comments.
1401 * doc/tm.texi (TARGET_SCHED_IS_COSTLY_DEPENDENCE): Update
1403 (TARGET_SCHED_ADJUST_COST_2): Remove documentation.
1405 * doc/rtl.texi (LOG_LINKS): Remove part about instruction scheduler.
1406 (REG_DEP_TRUE): Document.
1408 * config/ia64/ia64.c (ia64_adjust_cost_2): Rename to ia64_adjust_cost.
1409 Change signature to correspond to the targetm.sched.adjust_cost hook.
1410 Update use in TARGET_SCHED_ADJUST_COST_2.
1411 (TARGET_SCHED_ADJUST_COST_2): Rename to TARGET_SCHED_ADJUST_COST.
1412 (ia64_dependencies_evaluation_hook, ia64_dfa_new_cycle): Update to use
1413 new scheduler dependencies lists.
1414 (ia64_gen_check): Ditto.
1416 * config/mips/mips.c (vr4130_swap_insns_p): Update to use new scheduler
1419 * config/rs6000/rs6000.c (rs6000_is_costly_dependence): Change
1420 signature to correspond to the targetm.sched.is_costly_dependence hook.
1421 (is_costly_group): Update to use new scheduler dependencies lists.
1423 * config/spu/spu.c (spu_sched_adjust_cost): Use insn_cost () function
1424 instead of INSN_COST () macro.
1426 2007-02-01 Ian Lance Taylor <iant@google.com>
1428 * lower-subreg.c (resolve_clobber): Handle a subreg of a concatn.
1430 2007-02-01 Guy Martin <gmsoft@gentoo.org>
1432 * pa.md (tp_load): Correct mfctl instruction syntax.
1434 2007-02-01 Geoffrey Keating <geoffk@apple.com>
1436 * config/rs6000/rs6000.c (rs6000_stack_info): Correct
1437 altivec_padding_size calculation on AIX. Improve comment, add
1438 assert to verify that it's right.
1440 * config/rs6000/darwin.h (MD_UNWIND_SUPPORT): Don't define for
1443 2007-2-01 Seongbae Park <seongbae.park@gmail.com>
1446 * global.c (global_alloc): Add mising initialization of
1449 2007-02-01 Roger Sayle <roger@eyesopen.com>
1451 * alias.c (init_alias_analysis): Correct whitespace.
1452 * bb-reorder.c (fix_edges_for_rarely_executed_code,
1453 partition_hot_cold_basic_blocks): Likewise.
1454 * builtins.c (expand_builtin_printf, expand_builtin_fprintf,
1455 expand_builtin_sprintf, fold_builtin_carg, fold_builtin_sprintf,
1456 maybe_emit_sprintf_chk_warning, fold_builtin_sprintf_chk,
1457 fold_builtin_snprintf_chk, fold_builtin_printf,
1458 fold_builtin_fprintf, do_mpfr_ckconv, do_mpfr_arg1, do_mpfr_arg2,
1459 do_mpfr_arg3, do_mpfr_sincos): Likewise.
1460 * cfgcleanup.c (cleanup_cfg): Likewise.
1461 * cfgexpand.c (tree_expand_cfg): Likewise.
1462 * fold-const.c (fold_binary) <RDIV_EXPR>: Likewise.
1463 * function.c (get_next_funcdef_no): Likewise.
1464 * gengtype.c (main): Likewise.
1465 * genmodes.c (main): Likewise.
1466 * gcse.c (bypass_conditional_jumps, print_ldst_list): Likewise.
1467 * haifa-sched.c (schedule_block, extend_h_i_d): Likewise.
1468 * ifcvt.c (noce_emit_move_insn): Likewise.
1469 * modulo-sched.c (generate_prolog_epilog, sms_schedule_by_order):
1471 * stor-layout.c (get_best_mode): Likewise.
1472 * tree-ssa-loop-niter.c (get_val_for): Likewise.
1473 * tree-ssa-structalias.c (get_varinfo, get_varinfo_fc,
1474 scc_visit, do_ds_constraint, do_complex_constraint, label_visit,
1475 perform_var_substitution, solve_graph): Likewise.
1476 * tree-vrp.c (vrp_finalize): Likewise.
1478 2007-02-01 Ian Lance Taylor <iant@google.com>
1480 * lower-subreg.c (simplify_gen_subreg_concatn): If we ask for the
1481 high part of a paradoxical subreg, return a constant zero.
1483 2007-02-01 Zdenek Dvorak <dvorakz@suse.cz>
1485 * toplev.c (lang_dependent_init): Call init_set_costs.
1486 * loop-init.c (loop_optimizer_init): Do not call init_set_costs.
1488 2007-02-01 Richard Guenther <rguenther@suse.de>
1491 * fold-const.c (fold_negate_expr): Allow negating a
1492 constant if overflow does not change.
1494 2007-02-01 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1496 * doc/c-tree.texi (Expression trees): Improve markup.
1497 * doc/tm.texi (Register Classes, Addressing Modes)
1498 (Floating Point): Fix spacing after abbreviations. Fix some
1501 2007-02-01 Ben Elliston <bje@au.ibm.com>
1503 * doc/invoke.texi: Replace "bugfix" with "bug fix" throughout.
1504 * doc/contrib.texi: Likewise.
1505 * doc/install.texi: Likewise.
1507 2007-01-31 Richard Henderson <rth@redhat.com>
1508 Ian Lance Taylor <iant@google.com>
1510 * lower-subreg.c: New file.
1511 * rtl.def (CONCATN): Define.
1512 * passes.c (init_optimization_passes): Add pass_lower_subreg and
1514 * emit-rtl.c (update_reg_offset): New static function, broken out
1515 of gen_rtx_REG_offset.
1516 (gen_rtx_REG_offset): Call update_reg_offset.
1517 (gen_reg_rtx_offset): New function.
1518 * regclass.c: Revert patch of 2006-03-05, restoring
1520 (clear_reg_info_regno): New function.
1521 * dwarf2out.c (concatn_loc_descriptor): New static function.
1522 (loc_descriptor): Handle CONCATN.
1523 * common.opt (fsplit_wide_types): New option.
1524 * opts.c (decode_options): Set flag_split_wide_types when
1526 * timevar.def (TV_LOWER_SUBREG): Define.
1527 * rtl.h (gen_reg_rtx_offset): Declare.
1528 (reg_scan_update): Declare.
1529 * regs.h (clear_reg_info_regno): Declare.
1530 * tree-pass.h (pass_lower_subreg): Declare.
1531 (pass_lower_subreg2): Declare.
1532 * doc/invoke.texi (Option Summary): List -fno-split-wide-types.
1533 (Optimize Options): Add -fsplit-wide-types to -O1 list. Document
1535 * doc/rtl.texi (Regs and Memory): Document concat and concatn.
1536 * Makefile.in (OBJS-common): Add lower-subreg.o.
1537 (lower-subreg.o): New target.
1539 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1541 * config/sh/sh.h (HAVE_SECONDARY_RELOADS): Remove.
1543 2007-01-31 Anatoly Sokolov <aesok@post.ru>
1546 * config/avr/avr.c (DWARF2_ADDR_SIZE): Define.
1548 2007-01-31 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1551 * builtins.c (fold_builtin_sqrt): Use MPFR for constant args.
1553 2007-01-31 Zdenek Dvorak <dvorakz@suse.cz>
1555 * cfgloop.h: Include vec-prim.h.
1556 (enum li_flags): Remove LI_ONLY_OLD.
1557 (loop_iterator): Changed.
1558 (fel_next, fel_init): Iterate over loop tree.
1559 (FOR_EACH_LOOP_BREAK): New macro.
1560 * loop-unswitch.c (unswitch_loops): Do not pass LI_ONLY_OLD to
1562 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Ditto.
1563 * modulo-sched.c (sms_schedule): Ditto.
1564 * tree-vectorizer.c (vectorize_loops): Ditto.
1565 * doc/loop.texi: Update information on loop numbering and behavior of
1566 FOR_EACH_LOOP wrto new loops.
1567 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop,
1568 add_to_evolution_1): Test nestedness of loops instead of comparing
1570 * tree-chrec.c (chrec_fold_plus_poly_poly,
1571 chrec_fold_multiply_poly_poly, chrec_evaluate,
1572 hide_evolution_in_other_loops_than_loop, chrec_component_in_loop_num,
1573 reset_evolution_in_loop): Ditto.
1574 * Makefile.in (CFGLOOP_H): Add vecprim.h dependency.
1576 2007-01-31 Dirk Mueller <dmueller@suse.de>
1578 * c-common.c (warn_about_parentheses): Separate warning about
1579 un-parenthized sequence of comparison operators from the one
1580 which is supposed to warn about x <= y <= z.
1582 2007-01-31 Uros Bizjak <ubizjak@gmail.com>
1584 * optabs.h (enum optab_index): Add new OTI_isinf.
1585 (isinf_optab): Define corresponding macro.
1586 * optabs.c (init_optabs): Initialize isinf_optab.
1587 * genopinit.c (optabs): Implement isinf_optab using isinf?f2
1589 * builtins.c (mathfn_built_in): Handle BUILT_IN_ISINF{,F,L}.
1590 (expand_builtin_interclass_mathfn): Expand BUILT_IN_ISINF{,F,L}
1592 (expand_builtin): Expand BUILT_IN_ISINF{,F,L} using
1593 expand_builtin_interclass_mathfn.
1594 * reg_stack.c (subst_stack_regs_pat): Handle UNSPEC_FXAM.
1595 * config/i386/i386.md (UNSPEC_FXAM): New constant.
1596 (fxam<mode>2_i387): New insn pattern.
1597 (isinf<mode>2) New expander to implement isinf, isinff and isinfl
1598 built-in functions as x87 inline asm.
1600 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1602 * gcc/config/arm/unwind-arm.h (_sleb128_t, _uleb128_t): New.
1604 2007-01-30 Eric Christopher <echristo@apple.com>
1606 * config.gcc: Add geode.
1608 2007-01-31 Kazu Hirata <kazu@codesourcery.com>
1610 * cgraphunit.c, config/arm/arm.c, config/m68k/m68k.c,
1611 ipa-inline.c, tree-profile.c, tree-ssa-live.c,
1612 tree-ssa-math-opts.c, tree-ssanames.c, tree-vect-analyze.c,
1613 value-prof.c: Fix comment typos.
1615 2007-01-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1618 * doc/invoke.texi (Wpointer-arith): Document warning.
1620 2007-01-30 Janis Johnson <janis187@us.ibm.com>
1622 * doc/extend.texi (Decimal Floating Types): Remove decfloat.h from
1623 the list of discrepancies from the draft TR.
1625 2007-01-30 Dirk Mueller <dmueller@suse.de>
1628 * doc/invoke.texi (-Wreturn-type): Update description to
1631 2007-01-30 Richard Sandiford <richard@codesourcery.com>
1633 * cfgrtl.c (try_redirect_by_replacing_jump): Check only_sets_cc0_p.
1635 2007-01-30 Uros Bizjak <ubizjak@gmail.com>
1637 * builtins.c (expand_builtin_int_interclass_roundingfn): New function
1638 to handle optabs that operate on floating point input argument and
1639 output to integer output.
1640 (expand_builtin_mathfn) [BUILT_IN_ILOGB]: Move from here ...
1641 (expand_builtin_interclass_mathfn) [BUILT_IN_ILOGB]: ... to here.
1642 (expand_builtin): Expand BUILT_IN_ILOGB{,F,L} using
1643 expand_builtin_interclass_mathfn ().
1644 * config/i386/i386.md (fxtractxf3_i387): Rename from *fxtractxf3_i387.
1646 (ilogbxf2, ilogb<mode>2): New expanders to implement ilogb, ilogbf and
1647 ilogbl built-in functions as x87 intrinsics.
1649 2007-01-30 Richard Guenther <rguenther@suse.de>
1652 * dwarf2out.c (reference_to_unused): Query varpool if the
1653 variable was output.
1655 2007-01-30 Richard Guenther <rguenther@suse.de>
1658 * passes.c (execute_one_pass): Reset in_gimple_form to not
1659 confuse non-unit-at-a-time mode.
1661 2007-01-29 Roger Sayle <roger@eyesopen.com>
1662 Richard Guenther <rguenther@suse.de>
1664 * fold-const.c (round_up): Make HIGH an unsigned HOST_WIDE_INT to
1665 avoid undefined behaviour on overflow. Use force_fit_type_double
1666 to construct the constant with the specified TREE_OVERFLOW.
1668 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1670 * config/dfp-bit.c: Add parameterized support for fp exceptions.
1671 * config/dfp-bit.h: Ditto.
1673 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1675 * c-decl.c (pop_scope): Replace warnings with call to
1676 warn_for_unused_label.
1677 * c-common.h (warn_for_unused_label): Declare.
1678 * c-common.c (warn_for_unused_label): Define.
1680 2007-01-29 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
1682 * tree-optimize.c (update_inlined_to_pointers): Delete unused
1685 2007-01-29 Janis Johnson <janis187@us.ibm.com>
1687 * Makefile.in (USER_H): Remove decfloat.h.
1688 * ginclude/decfloat.h: Delete, moving contents to ...
1689 * ginclude/float.h: Add support for decimal floating point,
1690 guarded by __STDC_WANT_DEC_FP__.
1692 2007-01-29 Mike Stump <mrs@apple.com>
1694 * doc/gccint.texi (Top): Rename Loop Representation to Loop
1695 Analysis and Representation to resolve case insensitive conflict.
1696 * doc/loop.texi (Loop Analysis and Representation): Likewise.
1698 2007-01-28 Daniel Berlin <dberlin@dberlin.org>
1700 * tree.h (struct tree_memory_tag): Add aliases member.
1701 (MTAG_ALIASES): New macro.
1702 * tree-ssa-alias.c (alias_bitmap_obstack): New variable.
1703 (add_may_alias): Remove pointer-set. Update for may_aliases being
1705 (mark_aliases_call_clobbered): Update for may_aliases being a
1707 (compute_tag_properties): Ditto.
1708 (create_partition_for): Ditto.
1709 (compute_memory_partitions): Ditto.
1710 (dump_may_aliases_for): Ditto.
1711 (is_aliased_with): Ditto.
1712 (add_may_alias_for_new_tag): Ditto.
1713 (rewrite_alias_set_for): Rewrite for may_aliases being a bitmap.
1714 (compute_is_aliased): New function.
1715 (compute_may_aliases): Call compute_is_aliased).
1716 (init_alias_info): Initialize alias_bitmap_obstack.
1717 (union_alias_set_into): New function.
1718 (compute_flow_sensitive_aliasing): Use union_aliases_into.
1719 (have_common_aliases_p): Rewrite to take two bitmaps and use
1721 (compute_flow_insensitive_aliasing): Stop using pointer-sets.
1723 (finalize_ref_all_pointers): Update for add_may_alias changes.
1724 (new_type_alias): Ditto.
1725 * tree-flow-inline.h (may_aliases): Return a bitmap.
1726 * tree-dfa.c (dump_variable): Check for MTAG_P'ness.
1727 * tree-ssa.c (verify_flow_insensitive_alias_info): Update for
1728 may_aliases being a bitmap.
1729 * tree-flow.h (struct var_ann_d): Remove may_aliases member.
1730 may_aliases now returns a bitmap.
1731 * tree-ssa-structalias.c (merge_smts_into): Update for may_aliases
1733 * tree-ssa-operands.c (add_virtual_operand): Update for
1734 may_aliases being a bitmap.
1736 2007-01-29 Daniel Berlin <dberlin@dberlin.org>
1738 PR tree-optimization/30630
1739 * tree-ssa-structalias.c (do_complex_constraint): Mark correct
1740 variable as changed.
1742 2007-01-29 Simon Martin <simartin@users.sourceforge.net>
1745 * gimplify.c (gimplify_target_expr): Make sure that the TARGET_EXPR is
1746 expanded only once even if an error occurs.
1748 2007-01-29 Ben Elliston <bje@au.ibm.com>
1750 * gcov-io.h (__gcov_indirect_call_profiler): Declare.
1751 (__gcov_average_profiler): Likewise.
1752 (__gcov_ior_profiler): Likewise.
1753 (__gcov_merge_ior): Likewise.
1755 2007-01-28 Jan Hubicka <jh@suse.cz>
1757 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
1758 * value-prof.c (stringop_block_profile): Likewise.
1760 2007-01-28 Jan Hubicka <jh@suse.cz>
1762 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
1763 variant handling histograms; add wrapper.
1764 (clear_storage_via_libcall): Export.
1765 (emit_block_move_hints): Break out from ...; add histograms.
1766 (emit_block_move): ... this one.
1767 (clear_storage_hints): Break out from ...; add histograms.
1768 (clear_storage): ... this one.
1769 (set_storage_via_memset): Handle histogram.
1770 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
1771 (clear_storage_hints, clear_storage_via_libcall): Declare.
1772 (set_storage_via_setmem): Update prototype.
1773 * doc/md.texi (movmem, setmem): Document new arguments.
1775 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
1777 (stringop_block_profile): New global function.
1778 (tree_stringops_values_to_profile): Profile block size and alignment.
1779 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
1781 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
1782 (stringop_block_profile): Declare.
1783 * builtins.c: Include value-prof.h.
1784 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
1785 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
1786 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
1787 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
1788 * profile.c (instrument_values): Add new counters.
1789 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
1791 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
1792 (tree_init_edge_profiler): Build new profilers.
1793 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
1794 (pass_tree_profile): Add dump.
1795 (tree_profile_hooks): Update.
1796 * Makefile.in (LIBGCOV): Add new constants.
1797 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
1798 __gcov_ior_profiler): New.
1799 * i386.md (movmem/setmem expanders): Add new optional arguments.
1801 2007-01-28 David Edelsohn <edelsohn@gnu.org>
1803 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
1805 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1807 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
1808 (tree_unswitch_single_loop): Pass eni_size_weights to
1809 tree_num_loop_insns.
1810 * tree-ssa-loop-manip.c: Include tree-inline.h.
1811 (can_unroll_loop_p): Pass eni_size_weights to
1812 tree_num_loop_insns.
1813 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
1814 Pass eni_size_weights to estimate_num_insns.
1815 * tree.h (init_inline_once): Export.
1816 * toplev.c (backend_init): Call init_inline_once.
1817 * cgraphunit.c (cgraph_process_new_functions,
1818 cgraph_analyze_function): Pass eni_inlining_weights to
1820 * ipa-inline.c (compute_inline_parameters): Ditto.
1821 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
1822 to estimate_num_insns.
1823 (try_unroll_loop_completely): Pass eni_size_weights to
1824 tree_num_loop_insns.
1825 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
1826 ot estimate_num_insns.
1827 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
1828 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
1829 * tree-inline.c (eni_inlining_weights, eni_size_weights,
1830 eni_time_weights): New variables.
1831 (init_inline_once): Initialize them.
1832 (struct eni_data): Mew.
1833 (estimate_num_insns_1, estimate_num_insns): Use weights.
1834 * tree-inline.h (struct eni_weights_d): New.
1835 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
1836 (estimate_num_insns): Declaration changed.
1837 * cfgloop.h (tree_num_loop_insns): Declaration changed.
1838 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
1839 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
1841 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
1843 * tree-data-ref.c (conflict_fn): Assert that the number of affine
1844 relations in the conflict function is valid.
1846 2007-01-27 Ian Lance Taylor <iant@google.com>
1848 * common.opt: Add fstrict-overflow.
1849 * opts.c (decode_options): Set flag_strict_overflow if -O2.
1850 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
1851 (TYPE_OVERFLOW_UNDEFINED): Define.
1852 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
1854 * tree.h (TYPE_TRAP_SIGNED): Don't define.
1855 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
1856 (fold_negate_expr): Likewise.
1857 (make_range): Likewise.
1858 (extract_muldiv_1): Likewise.
1859 (maybe_canonicalize_comparison): Likewise.
1860 (fold_comparison): Likewise.
1861 (fold_binary): Likewise.
1862 (tree_expr_nonnegative_p): Likewise.
1863 (tree_expr_nonzero_p): Likewise.
1864 * tree-vrp.c (compare_values): Likewise.
1865 (extract_range_from_binary_expr): Likewise.
1866 (extract_range_from_unary_expr): Likewise.
1867 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
1869 (nowrap_type_p): Likewise.
1870 * tree-scalar-evolution.c (simple_iv): Likewise.
1871 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
1872 (build_range_check): Likewise.
1873 (extract_muldiv_1): Likewise.
1874 (fold_comparison): Likewise.
1875 * tree-vrp.c (vrp_int_const_binop): Likewise.
1876 (extract_range_from_unary_expr): Likewise.
1877 * convert.c (convert_to_integer): Likewise.
1878 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
1879 (fold_comparison): Likewise.
1880 (fold_binary): Likewise.
1881 * optabs.c (optab_for_tree_code): Likewise.
1882 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
1883 * simplify-rtx.c (simplify_const_relational_operation): Check
1884 flag_strict_overflow and flag_trapv.
1885 (simplify_const_relational_operation): Likewise.
1886 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
1887 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
1890 2007-01-27 Roger Sayle <roger@eyesopen.com>
1892 * tree.c (tree_fold_gcd): Delete.
1893 * tree.h (tree_fold_gcd): Remove prototype.
1894 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
1895 test whether one constant integer is a multiple of another. Instead
1896 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
1897 * fold-const.c (multiple_of_p): We've determined both TOP and
1898 BOTTOM are integer constants so we can call int_const_binop directly
1899 instead of the more generic const_binop.
1901 2007-01-27 Roger Sayle <roger@eyesopen.com>
1903 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
1904 1*X check that the constant hasn't overflowed, to preserve the
1906 (round_up): Provide an efficient implementation when rouding-up an
1907 INTEGER_CST to a power-of-two.
1909 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1911 * doc/sourcebuild.texi: Add comma for clarity.
1912 * doc/extend.texi: Fix some typos.
1913 * doc/passes.texi: Likewise.
1914 * doc/cppinternals.texi: Likewise.
1915 * doc/c-tree.texi: Likewise.
1916 * doc/tree-ssa.texi: Likewise.
1917 * doc/install.texi: Likewise.
1919 2007-01-27 Jan Hubicka <jh@suse.cz>
1921 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
1922 (pass_sra): Do not require alias information.
1923 * passes.c (init_optimization_passes): Add SRA
1925 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
1927 * tracer.c (rest_of_handle_tracer): We already cleaned
1928 up the CFG in tracer() so don't do it here again.
1929 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
1930 cleanup_cfg here, either. And don't call renumber_insns.
1932 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
1933 and BB_HEAD are in the insn stream here. Instead make sure
1934 that BB_INSN is valid on all insns. Also, do check here that
1935 there are no pending branch predictions...
1936 (rtl_verify_flow_info): ...instead of doing it here. Checks
1937 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
1940 2007-01-26 Roger Sayle <roger@eyesopen.com>
1942 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
1943 function to simplify/factorize operand order canonicalization.
1944 (ix86_fixup_binary_operands): Reorganize using the above function.
1945 (ix86_binary_operator_ok): Likewise.
1947 2007-01-27 Jakub Jelinek <jakub@redhat.com>
1949 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
1950 file scope from optimize_attrs.
1951 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
1953 (optimize_attrs): Clear insn_code_values after freeing it.
1955 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1957 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
1958 a parts.base assignment.
1960 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
1962 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
1963 (compute_subscript_distance, initialize_data_dependence_relation,
1964 finalize_ddr_dependent, analyze_ziv_subscript,
1965 analyze_siv_subscript_cst_affine,
1966 compute_overlap_steps_for_affine_univar,
1967 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
1968 analyze_siv_subscript, analyze_miv_subscript,
1969 analyze_overlapping_iterations, subscript_dependence_tester_1,
1970 compute_self_dependence, free_dependence_relation): Work
1971 with affine_fn instead of chrecs.
1972 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
1973 common_affine_function, affine_function_base,
1974 affine_function_constant_p, affine_fn_op, affine_fn_plus,
1975 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
1976 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
1977 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
1978 (all_chrecs_equal_p): Removed.
1979 * tree-data-ref.h (affine_fn, conflict_function): New types.
1980 (struct subscript): Change type of conflicting_iterations_in_a
1981 and conflicting_iterations_in_b.
1983 2007-01-26 Steve Ellcey <sje@cup.hp.com>
1986 * config/pa/pa.h (TARGET_HPUX_11): New.
1987 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
1988 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
1990 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
1991 Richard Guenther <rguenther@suse.de>
1993 * tree-ssa-structalias.c (solve_graph): Handle case
1994 we merged the variable to another.
1996 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1998 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
2001 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
2002 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
2004 2007-01-25 DJ Delorie <dj@redhat.com>
2006 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
2007 allow changes to modes which don't fit in those registers.
2009 * reload1.c (choose_reload_regs): Check for invalid subregs before
2010 computing their locations, not after.
2012 2007-01-25 Geoffrey Keating <geoffk@apple.com>
2015 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
2016 save Altivec registers in an eh_return function.
2017 (compute_vrsave_mask): Likewise.
2018 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
2019 for saving Altivec registers.
2020 (rs6000_emit_prologue): Don't allocate stack twice in
2021 eh_return function. Correct expected value of altivec_save_offset
2022 when using save_world. Describe save of R0 to stack when using
2023 save_world. Describe stack pointer adjustment when using
2024 save_world. Remove duplicated eh_return parameter register saving.
2025 Update sp_offset variable after save_world.
2026 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
2028 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
2029 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
2032 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2034 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
2036 2007-01-25 Steve Ellcey <sje@cup.hp.com>
2039 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
2040 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
2042 2007-01-25 Richard Guenther <rguenther@suse.de>
2044 * doc/invoke.texi (-Wcoverage-mismatch): Document.
2045 * common.opt (-Wcoverage-mismatch): New warning option.
2046 * coverage.c (get_coverage_counts): Ignore coverage mismatch
2047 if -Wcoverage-mismatch is given.
2049 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
2051 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
2052 (ipcp_driver): Change to static definition.
2053 Add dumping of the ifunctions.
2054 (constant_val_insert): Remove unused parameter. Support for SSA.
2055 (ipcp_propagate_const): Support for SSA.
2056 (ipcp_profile_bb_print): Print only analyzed nodes.
2057 (ipcp_replace_map_create): Remove support for Fortran constant
2059 * ipa-prop.c (ipa_method_modify_stmt,
2060 ipa_callsite_compute_param): Support for SSA.
2061 * ipa-prop.h (ipcp_driver): Remove declaration.
2062 (IS_VALID_TREE_MAP_INDEX): Add define.
2064 2007-01-24 Geoffrey Keating <geoffk@apple.com>
2066 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
2068 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2071 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
2072 constant argument to cabs and do it without checking for
2073 -funsafe-math-optimizations.
2075 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
2077 * c-common.h (RID_FIRST_CXX0X): New.
2078 (RID_LAST_CXX0X): New.
2079 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
2081 * c.opt (Wc++0x-compat): New.
2082 * doc/invoke.texi (-Wc++0x-compat): Document.
2084 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2086 * builtins.c (fold_builtin_carg): New.
2087 (fold_builtin_1): Use it.
2089 2007-01-24 Jan Hubicka <jh@suse.cz>
2091 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
2093 * ipa-inline.c (initial_insns, max_insns): Delete.
2094 (compute_max_insns): New function.
2095 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
2096 of insns as base for code growth.
2097 (cgraph_decide_inlining): Make initial_insns local; do not compute
2099 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
2100 * doc/invoke.texi (inline-unit-growth): Update docs.
2102 2007-01-24 Jakub Jelinek <jakub@redhat.com>
2104 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
2105 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
2106 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
2107 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
2108 for CPUs that have PTA_CX16 set.
2111 * gimplify.c (omp_check_private): New function.
2112 (gimplify_scan_omp_clauses): Use it for
2113 firstprivate/lastprivate/reduction.
2116 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
2117 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
2120 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
2121 If lower_lastprivate_clauses emits some statements, append them
2122 to dlist rather than body_p and to body_p append an initializer.
2123 (lower_omp_for): Adjust caller.
2125 2007-01-24 Steve Ellcey <sje@cup.hp.com>
2127 * target.h (globalize_decl_name): New.
2128 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2129 * output.h (default_globalize_decl_name): New.
2130 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
2132 (globalize_decl): Ditto.
2133 (default_globalize_decl_name): New.
2134 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
2135 (ia64_handle_version_id_attribute): New.
2136 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
2137 (ia64_asm_output_external): Use globalize_decl_name instead
2139 * doc/extend.texi (version_id): New pragma.
2140 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
2142 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2144 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
2145 _uleb128_t and _Unwind_SWord with _sleb128_t.
2146 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
2147 uw_frame_state_for, uw_update_context_1): Likewise.
2148 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
2149 * unwind-pe.h (read_uleb128, read_sleb128,
2150 read_encoded_value_with_base): Likewise.
2151 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
2153 2007-01-24 Richard Guenther <rguenther@suse.de>
2155 * builtins.c (expand_builtin_cexpi): Get the fndecl
2156 for cexp in the correct way.
2158 2007-01-24 Jan Hubicka <jh@suse.cz>
2160 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
2162 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
2164 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
2165 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
2166 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
2167 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
2168 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
2169 __SIZEOF_PTRDIFF_T__.
2170 * doc/cpp.texi: Documentation for the new builtin macros added.
2172 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
2174 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
2175 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
2176 instead of operands[2] to avoid access past the end of array.
2178 2007-01-24 Richard Sandiford <richard@codesourcery.com>
2180 * reload1.c (emit_reload_insns): Pass the reload register
2181 for a non-spill output reload through forget_old_reloads_1.
2183 2007-01-23 Joseph Myers <joseph@codesourcery.com>
2185 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
2186 _SOFT_DOUBLE if doubles use software floating-point.
2187 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
2188 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2189 * config/rs6000/darwin-ldouble.c: Also compile functions for
2190 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
2191 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
2192 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
2193 commented-out long double override.
2194 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
2195 * config/rs6000/eabispe.h: Likewise.
2196 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
2197 long double for non-SPE.
2198 (rs6000_handle_option): Likewise.
2199 (invalid_e500_subreg): Disallow more subregs involding DImode,
2200 DFmode, TImode or TFmode.
2201 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
2203 (legitimate_lo_sum_address_p): Also check for TFmode for E500
2205 (rs6000_legitimize_address): Also handle TFmode for E500 double.
2206 (rs6000_legitimize_reload_address): Also handle TFmode for E500
2208 (rs6000_legitimate_address): Also check for TFmode for E500
2210 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
2211 (spe_build_register_parallel): Handle TFmode and TCmode.
2212 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
2214 (function_arg): Handle TFmode and TCmode for E500 double.
2215 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
2217 (print_operand): Handle TFmode and TImode for %y.
2218 (rs6000_generate_compare): Handle TFmode comparisons for E500
2220 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
2221 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
2222 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
2223 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
2224 TFmode for E500 double.
2225 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
2226 (floatsidf2): Enable for E500 double.
2227 (movtf_softfloat): Use rs6000_nonimmediate_operand.
2228 (extenddftf2): Change to extenddftf2_fprs.
2229 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
2230 depending on TARGET_E500_DOUBLE.
2231 (extendsftf2): Enable for E500 double.
2232 (trunctfdf2): Enable for E500 double.
2233 (trunctfsf2): Change to trunctfsf2_fprs.
2234 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
2235 depending on TARGET_E500_DOUBLE.
2236 (floatsitf2): Enable for E500 double.
2237 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
2238 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
2239 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
2240 (negtf2): Change to negtf2_internal.
2241 (negtf2): New expander.
2242 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
2243 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
2244 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
2245 (movdi_internal32): Use rs6000_nonimmediate_operand.
2246 (unnamed splitter): Likewise.
2247 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
2248 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
2249 (SPE64TF, DITI): New mode macros.
2250 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
2253 (frob_<mode>_di_2): New.
2254 (frob_tf_di_8_2): New.
2255 (frob_di_df): Change to frob_di_<mode>; allow more modes.
2257 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
2259 (frob_ti_<mode>_8_2): New.
2260 (frob_ti_tf_2): New.
2261 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
2262 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
2264 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
2265 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
2266 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
2267 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
2268 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
2269 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
2272 2007-01-23 Ian Lance Taylor <iant@google.com>
2274 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
2276 (OBJS-archive): Reformat, alphabetize.
2277 (OBJS): Change out_object_file to OBJS-md.
2279 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
2281 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
2282 JUMP_P and CALL_P predicates where applicable.
2283 * config/i386/i386.c: Ditto.
2284 * config/i386/i386.md: Ditto.
2285 * config/i386/mmx.md: Ditto.
2286 * config/i386/predicates.md: Ditto.
2288 2007-01-22 Andreas Schwab <schwab@suse.de>
2290 * config/m68k/m68k.h: Fix comment.
2292 2007-01-22 Jan Hubicka <jh@suse.cz>
2294 * passes.c (init_optimization_passes): Do not rerun
2295 pass_early_warn_uninitialized.
2297 2007-01-22 Richard Guenther <rguenther@suse.de>
2299 PR tree-optimization/30038
2300 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
2302 (execute_cse_sincos_1): Likewise.
2303 (execute_cse_sincos): Likewise.
2304 (gate_cse_sincos): Likewise.
2305 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
2306 the cexpi() canonicalization of sincos().
2307 * tree-pass.h (pass_cse_sincos): Declare.
2308 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
2310 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
2312 PR rtl-optimization/29329
2313 * combine.c (replaced_rhs_insn): Rename to i2mod.
2314 (replaced_rhs_value): Rename to i2mod_new_rhs.
2315 (i2mod_old_rhs): New global variable.
2316 (combine_instructions): Adjust for above change. Save a copy of
2317 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
2318 note are substituted in the second instruction.
2319 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
2320 ditch the note if it pertains to the second eliminated register
2321 and this register is mentioned in i2mod_old_rhs.
2324 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
2326 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
2327 the insn to which the note was originally attached.
2329 2007-01-21 Jan Hubicka <jh@suse.cz>
2331 * ipa-inline.c (inlining_mode): Comment, move up.
2332 (cgraph_decide_inlining_incrementally): Do not perform inlining
2333 itself; fix handling of flattening of self recursive functions.
2334 (cgraph_find_cycles): Remove.
2335 (cgraph_flatten_node): Remove.
2336 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
2337 (try_inline): New function.
2338 (cgraph_early_inlining): Update call of
2339 cgraph_decide_inlining_incrementally. Apply inlining here.
2340 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
2342 2007-01-21 Dirk Mueller <dmueller@suse.de>
2345 * tree-vrp.c (check_array_bounds): do not warn
2346 about ADDR_EXPR's of ARRAY_REF's which are immediately
2347 used in binary expressions.
2349 2007-01-21 Ira Rosen <irar@il.ibm.com>
2351 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
2352 read_write_dep and macros for its access.
2353 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2354 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
2355 argument, call vect_check_interleaving for every independent pair of
2356 data-refs. Mark loads that access the same memory location as a store
2358 (vect_check_dependences): Remove.
2359 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
2360 call, fix the call to vect_analyze_data_ref_dependence.
2361 (vect_analyze_data_ref_access): For statements that access the same
2362 data-ref, check that they are not stores; for loads, check that there
2363 is no store that access the same location.
2365 2007-01-20 Roger Sayle <roger@eyesopen.com>
2366 Joseph Myers <joseph@codesourcery.com>
2368 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
2369 (-fdump-rtl-stack): Likewise.
2370 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
2371 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
2372 (-mpt-fixed): Use @minus{} for minus sign.
2373 (Using Precompiled Headers): Fix under/overfull hbox.
2375 2007-01-20 Jan Hubicka <jh@suse.cz>
2377 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
2378 place within annotation.
2380 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2382 * pa.c (output_move_double): Change array size of xoperands to 4.
2384 2007-01-20 Jan Hubicka <jh@suse.cz>
2386 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
2389 2007-01-19 Ian Lance Taylor <iant@google.com>
2391 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
2392 constant to unsigned HOST_WIDE_INT before negating.
2394 2007-01-19 Ian Lance Taylor <iant@google.com>
2396 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
2398 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
2399 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2400 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
2401 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
2402 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
2403 (realloc_vdef, realloc_vuse): Change second parameter to
2405 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
2407 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
2408 unsigned. Change x and lim locals to unsigned.
2409 (realloc_vdef, realloc_vuse): Change num_elem parameter to
2411 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
2412 (copy_virtual_operands): Change i and n locals to unsigned.
2414 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2417 * builtins.c (fold_builtin_1): Handle builtin fdim.
2419 2007-01-20 Jan Hubicka <jh@suse.cz>
2421 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
2422 * tree-ssa-structalias.c (compute_points_to_sets): Do call
2423 init_alias_heapvars.
2424 (init_alias_heapvars): Initialize only when not already initialized.
2425 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
2427 2007-01-19 Roger Sayle <roger@eyesopen.com>
2429 * common.opt (fsigned-zeros): New command line option.
2430 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
2431 of flag_unsafe_math_optimizations.
2432 * opts.c (set_fast_math_flags): The -ffast-math command line option
2433 implies -fno-signed-zeros.
2434 (fast_math_flags_set_p): Likewise.
2436 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
2437 the documentation of -ffast-math appropriately. Wrap long lines.
2439 2007-01-19 Steve Ellcey <sje@cup.hp.com>
2441 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
2442 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
2444 2007-01-19 Tomas Bily <tbily@suse.cz>
2446 * cgraphunit.c (cgraph_finalize_function): Updating of pid
2448 (tree_init_ic_make_global_vars): New function
2449 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
2450 (tree_gen_ic_profiler): New function
2451 (tree_gen_ic_func_profiler): New function
2452 (tree_profiling): Added calling of tree_gen_ic_func_profiler
2453 (tree_profile_hooks): Added hook for indirec/virtual calls
2454 * value-prof.c (tree_find_values_to_profile): New case for
2456 (tree_values_to_profile): Call for determining indirect/virtual
2458 (tree_indirect_call_to_profile): New function
2459 (tree_ic_transform): New function
2460 (tree_ic): New function
2461 (find_func_by_pid): New function
2462 (init_pid_map): New function
2463 (tree_value_profile_transformations): Added check for
2464 indirect/virtual call transformation
2465 * value-prof.h (enum hist_type): New counter type for
2466 indirect/virtual calls
2467 (profile_hooks): Added new hook for profiling indirect/virtual
2469 * profile.c (instrument_values): New case for indirect/virtual
2471 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
2472 (GCOV_COUNTER_V_INDIR): New counter type
2473 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
2474 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
2476 * cgraph.c: Definition of cgraph_max_pid
2477 (cgraph_create_node): Default init of pid attribute
2478 * cgraph.h: Declaration of cgraph_max_pid
2479 (struct cgraph_node): Added pid attribute
2480 * libgcov.c (__gcov_indirect_call_profiler): New function
2481 (__gcov_one_value_profiler_body): New function
2482 (__gcov_one_value_profiler): Body was moved to
2483 __gcov_one_value_profiler_body and calls it
2485 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
2487 * doc/gty.texi (Options): Document the mark_hook option to GTY.
2488 * gengtype.c (write_types_data, write_func_for_structure,
2489 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
2490 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
2491 (walk_type, write_func_for_structure): Generate the mark_hook if
2494 2007-01-19 Jan Hubicka <jh@suse.cz>
2496 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
2497 'early' argument take inlining mode argument specifying whether to
2498 inline for size/speeed or all functions; add support for flattening;
2500 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
2502 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2505 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
2506 the name of the declared identifier.
2508 2007-01-19 Dirk Mueller <dmueller@suse.de>
2510 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
2511 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
2512 duplicate evaluation.
2514 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2516 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
2517 Macroize expander using X87MODEF12 mode macro. Extend operand 1
2518 to XFMode, use acosxf2 and truncate result to requested mode.
2519 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
2520 (asin<mode>2): Similarly, with asin expanders.
2521 (*fscalexf4_i387): Rename from *fscalexf4.
2522 (expNcorexf3): New expander.
2523 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
2524 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
2525 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
2526 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
2527 patterns for SSE math.
2528 (exp10<mode>2): Similarly, with exp10 expanders.
2529 (exp2<mode>2): Similarly, with exp2 expanders.
2530 (expm1<mode>2): Similarly, with expm1 expanders.
2531 (ldexp<mode>3): Similarly, with ldexp expanders.
2532 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
2533 Use gen_truncxf<mode>2_i387_noop to truncate result.
2535 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2537 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
2538 for ColdFire targets.
2540 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
2541 Richard Sandiford <richard@codesourcery.com>
2543 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
2544 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
2545 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
2546 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
2547 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2548 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
2550 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2552 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
2553 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
2554 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
2555 (*adddi_dilshr32_cf): New pattern.
2556 (adddi3, subdi3): Remove first alternatives.
2558 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2560 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
2561 implemented using move.l, do not use its cc result for floating-point
2564 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2566 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
2567 from accepting 's' constraints if flag_pic.
2569 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2571 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
2572 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
2573 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
2574 condition from TARGET_68881 to TARGET_HARD_FLOAT.
2576 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
2578 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
2579 ColdFire alternatives.
2580 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
2581 * config/m68k/m68k.md (clzsi2): Define for ColdFire
2582 architectures that support ff1 instruction.
2584 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2585 Julian Brown <julian@codesourcery.com>
2587 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
2588 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
2589 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
2592 2007-01-19 Richard Sandiford <richard@codesourcery.com>
2594 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
2596 (output_move_const_into_data_reg, output_move_simode_const): Delete.
2597 * config/m68k/m68k.c (const_method, const_int_cost): Take a
2598 HOST_WIDE_INT instead of an rtx.
2599 (m68k_rtx_costs): Update call accordingly.
2600 (output_move_const_into_data_reg): Likewise. Fix formatting.
2601 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
2603 (output_move_simode_const): Update calls after above changes.
2604 Rework to use automatic variables and predicates like MEM_P.
2605 * config/m68k/m68k.md (pushexthisi_const): Update call to
2608 2007-01-19 Dirk Mueller <dmueller@suse.de>
2610 * tree-ssa-alias.c (perform_var_substitution): Fix typo
2613 2007-01-19 Richard Guenther <rguenther@suse.de>
2615 * builtins.c (expand_builtin_cexpi): Fall back to expanding
2616 via cexp in case sincos is not available.
2618 2007-01-19 Richard Guenther <rguenther@suse.de>
2620 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
2621 * defaults.h (TARGET_HAS_SINCOS): Default to off.
2622 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
2623 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
2624 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
2625 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
2626 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
2627 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
2629 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
2631 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
2633 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
2634 (atan2xf3): Directly generate RTL pattern.
2635 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
2636 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2637 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2638 disable patterns for SSE math.
2639 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
2640 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
2641 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
2642 disable patterns for SSE math.
2644 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
2646 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
2648 2007-01-18 Roger Sayle <roger@eyesopen.com>
2650 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
2651 VIEW_CONVERT_EXPR to the same type as it's operand.
2653 2007-01-18 David Edelsohn <edelsohn@gnu.org>
2655 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
2656 configured for long double 128.
2658 2007-01-18 Mike Stump <mrs@apple.com>
2660 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
2663 2007-01-18 Michael Meissner <michael.meissner@amd.com>
2665 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
2668 2007-01-18 Jan Hubicka <jh@suse.cz>
2670 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
2671 greater than NUM_VOP_FREE_BUCKETS.
2673 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
2675 * tree-ssa-structalias.c: Update comments.
2676 (ptabitmap_obstack): Removed.
2678 (oldpta_obstack): Ditto.
2679 (stats): Add a few members.
2680 (struct variable_info): Remove node, complex, address_taken, and
2681 indirect_target members. Add oldsolution member.
2682 (new_var_info): Do not initialize removed members.
2683 (constraint_expr_type): Remove INCLUDES.
2684 (constraint_graph): Add size, implicit_preds, rep,
2685 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
2686 (FIRST_REF_NODE): New macro.
2687 (LAST_REF_NODE): Ditto.
2688 (FIRST_ADDR_NODE): Ditto.
2689 (find): New function.
2691 (dump_constraint): Do not handle INCLUDES.
2692 (insert_into_complex): Do not insert duplicate constraints.
2693 (condense_varmap_nodes): Renamed and rewritten into ...
2694 (merge_node_constraints): This. Also fix bug in handling of
2695 offseted copy constraints.
2696 (clear_edges_for_node): No longer need to deal with preds at all,
2697 or removing associated preds/succs.
2698 (merge_graph_nodes): Deal with indirect_cycles.
2699 Don't deal with predecessors.
2700 (add_implicit_graph_edge): New function.
2701 (add_pred_graph_edge): Ditto.
2702 (add_graph_edge): Don't deal with predecessors.
2703 (build_constraint_graph): Removed.
2704 (build_pred_graph): New function.
2705 (build_succ_graph): Ditto.
2706 (struct scc_info): Removed in_component. Added roots, dfs, and
2707 node_mapping. Remove visited_index, unification_queue.
2708 (scc_visit): Deal with union-find we do now.
2709 Deal with cycles with REF nodes.
2710 (collapse_nodes): Renamed and rewritten to ...
2711 (unify_nodes): This.
2712 (process_unification_queue): Removed.
2713 (topo_visit): Cleanup
2714 (do_da_constraint): Use find.
2715 (do_sd_constraint): Ditto.
2716 (do_ds_constraint): Ditto.
2717 (do_complex_constraint): Ditto.
2718 (init_scc_info): Update for removed and added members.
2719 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
2720 (find_indirect_cycles): This.
2721 (equivalence_class): New variable.
2722 (label_visit): New function.
2723 (perform_variable_substitution): Rewritten.
2724 (free_var_substitution_info): New function.
2725 (find_equivalent_node): Ditto.
2726 (move_complex_constraints): Ditto.
2727 (eliminate_indirect_cycles): Ditto.
2728 (solve_graph): Only propagate changed bits.
2729 Use indirect cycle elimination.
2731 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
2732 (tree_id_eq): Renamed to ...
2733 (tree_vi_eq): This. Update for member change
2734 (insert_id_for_tree): Renamed and rewritten to ...
2735 (insert_vi_for_tree): This.
2736 (lookup_id_for_tree): Renamed and rewritten to ...
2737 (lookup_vi_for_tree): This.
2738 (get_id_for_tree): Renamed and rewritten to ...
2739 (get_vi_for_tree): Ditto.
2740 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
2741 (process_constraint): Don't handle INCLUDES.
2742 Remove special ADDRESSOF case.
2743 (find_func_aliases): Rewrite to use vi functions instead of id
2745 (create_function_info_for): Ditto.
2746 (create_variable_info_for): Ditto.
2747 (intra_create_variable_infos): Ditto.
2748 (merge_smts_into): Ditto.
2749 (find_what_p_points_to): Ditto.
2750 (init_base_vars): Ditto.
2751 (init_alias_vars): Ditto.
2752 (remove_preds_and_fake_succs): New function.
2753 (dump_sa_points_to_info): Dump new stats.
2754 (dump_solution_for_var): Use find.
2755 (set_used_smts): Fix formatting.
2756 (compute_points_to_sets): Updated for new functions.
2757 (ipa_pta_execute): Ditto.
2759 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2760 Richard Sandiford <richard@codesourcery.com>
2762 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
2764 * calls.c (expand_call): If the return value is a PARALLEL,
2765 extract its first member.
2766 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
2767 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2768 use FUNCTION_EXTRA_EPILOGUE.
2769 (m68k_function_value): Return a PARALLEL if the return value
2770 is of a pointer type.
2771 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
2772 (FUNCTION_EXTRA_EPILOGUE): Remove.
2773 * config/m68k/m68k.md (D0_REG): New constant.
2775 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2777 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
2778 output a NOP for empty epilogues.
2780 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2782 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
2783 before function. Extend register save check to include all
2784 registers, not just integer ones.
2786 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2788 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
2789 (m68k_use_return_insn): ...this.
2790 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
2791 * config/m68k/m68k.c (use_return_insn): Rename to...
2792 (m68k_use_return_insn): ...this.
2793 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
2796 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2798 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
2799 nearest even, fix denormal rounding overflow.
2801 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2803 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
2804 f%$smove and f%$move.
2805 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
2806 (extendsfdf2_cf): Use fdmove instead of f%&move.
2807 (truncdfsf2_cf): Use fsmove instead of f%$smove.
2808 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
2810 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2812 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
2815 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2817 * real.h (real_format): Add a canonical_nan_lsbs_set field.
2818 (coldfire_single_format): Declare.
2819 (coldfire_double_format): Likewise.
2820 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
2821 of qnan_msb_set to determine the lower bits of a canonical
2823 (encode_ieee_double): Likewise.
2824 (encode_ieee_quad): Likewise.
2825 (ieee_single_format): Initialize canonical_nan_lsbs_set.
2826 (mips_single_format): Likewise.
2827 (ieee_double_format): Likewise.
2828 (mips_double_format): Likewise.
2829 (ieee_extended_motorola_format): Likewise.
2830 (ieee_extended_intel_96_format): Likewise.
2831 (ieee_extended_intel_128_format): Likewise.
2832 (ieee_extended_intel_96_round_53_format): Likewise.
2833 (ibm_extended_format): Likewise.
2834 (mips_extended_format): Likewise.
2835 (ieee_quad_format): Likewise.
2836 (mips_quad_format): Likewise.
2837 (vax_f_format): Likewise.
2838 (vax_d_format): Likewise.
2839 (vax_g_format): Likewise.
2840 (i370_single_format): Likewise.
2841 (i370_double_format): Likewise.
2842 (decimal_single_format): Likewise.
2843 (decimal_double_format): Likewise.
2844 (decimal_quad_format): Likewise.
2845 (c4x_single_format): Likewise.
2846 (c4x_extended_format): Likewise.
2847 (real_internal_format): Likewise.
2848 (coldfire_single_format): New real_format.
2849 (coldfire_double_format): Likewise.
2850 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
2851 canonical_nan_lsbs_set.
2852 (pdp11_d_format): Likewise.
2853 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
2854 if TARGET_COLDFIRE_CPU.
2856 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2858 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
2859 (output_call, m68k_legitimize_call_address): Declare.
2860 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
2861 parenthesees. Add support for a 'W' constraint.
2862 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
2863 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
2864 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
2865 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
2867 (override_options): Initialize them. Do not set flag_no_function_cse
2868 for TARGET_ID_SHARED_LIBRARY.
2869 (m68k_output_pic_call): Delete.
2870 (m68k_legitimize_call_address): New function.
2871 (print_operand): Remove the %o prefix. Handle the %p prefix.
2872 (output_call): New function.
2873 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
2874 target address from the GOT if symbolic jumps are not allowed.
2875 * config/m68k/m68k.md (call, general_operand): Do not set
2876 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
2877 Merge separate flag_pic and !flag_pic define_insns into...
2878 (*call, *call_value): ...these new patterns. Match the address
2879 rather than the containing MEM and require it to be a call_operand.
2880 Use output_call to generate the asm template.
2881 * config/m68k/predicates.md (const_call_operand): New predicate.
2882 (call_operand): Likewise.
2884 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2886 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
2887 (STACK_GROWS_DOWNWARD): Define to 1.
2888 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
2889 comments, emphasizing that these values are only defaults.
2890 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
2891 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
2893 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2895 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
2897 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
2898 Richard Sandiford <richard@codesourcery.com>
2900 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
2901 (FP_REGNO_P): New macros.
2902 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
2903 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
2904 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
2905 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
2907 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2909 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2910 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
2912 * config/m68k/t-floatlib: New file, extracting common code from...
2913 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
2914 * config/m68k/t-uclinux: Here.
2915 * config/m68k/fpgnulib.c: Do not compile extendeed precision
2916 routines on ColdFire targets.
2918 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
2920 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
2921 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
2922 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
2923 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
2924 explicitly set MOTOROLA to 1.
2925 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
2927 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2928 Nathan Sidwell <nathan@codesourcery.com>
2930 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
2931 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
2932 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
2933 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
2934 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
2935 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
2936 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
2937 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
2938 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
2939 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2940 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
2941 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
2942 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
2943 (DBX_REGISTER_NUMBER): Delete.
2944 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
2945 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
2946 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
2948 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2950 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
2951 * config/flat.h: New file.
2952 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
2954 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
2955 indirect references for -msep-data or -mid-shared-library.
2956 Do not use PC-relative code addresses either.
2957 * config/m68k/m68k.c (override_options): Restrict -fPIC error
2959 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
2960 for shared libraries and crt1.o for executables. Use crti.o and
2962 (ENDFILE_SPEC): Use crtend.o and crtn.o.
2963 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
2964 Do not add -elf2flt or -shared-lib-id options here.
2965 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
2966 Pass -shared-lib-id if -mid-shared-library, taking the library
2967 identifier from -mshared-library-id if given, otherwise
2969 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
2970 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
2971 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
2972 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
2973 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
2974 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
2976 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
2977 jump sequence for ISA A and ISA A+.
2978 (PICJUMP): Likewise.
2980 2007-01-18 Richard Sandiford <richard@codesourcery.com>
2982 * config.gcc (m68k-*-uclinux*): Base the port on the common
2983 and m68k GNU/Linux files rather than on the generic ELF ones.
2984 * config/m68k/uclinux.h (TARGET_VERSION): Override.
2985 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
2987 2007-01-18 Julian Brown <julian@codesourcery.com>
2988 Richard Sandiford <richard@codesourcery.com>
2990 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
2991 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2992 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
2994 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
2996 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2998 2007-01-18 Richard Sandiford <richard@codesourcery.com>
3000 * doc/install.texi: Document m68k-uclinuxoldabi.
3001 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
3002 * config/m68k/uclinux-oldabi.h: New file, copied from
3003 config/m68k/uclinux.h.
3005 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
3007 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
3008 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
3009 (m32r_compute_frame_size): Use unsigned for regno.
3010 (m32r_reload_lr): Use gen_frame_mem.
3012 (m32r_output_function_epilogue): Don't output the function epilogue
3014 (m32r_expand_epilogue): New.
3015 (direct_return): Return false if the function has the interrupt
3017 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
3018 * config/m32r/m32r.md (epilogue): New expander.
3019 (return_lr, return_rte): New insns.
3020 (return): Make it expander.
3021 (return_normal): New expander.
3023 2007-01-18 Josh Conner <jconner@apple.com>
3026 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
3027 support for UNLE, UNLT, UNGE, and UNGT.
3029 2007-01-18 Jan Hubicka <jh@suse.cz>
3031 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
3032 * tree-tailcall (add_virtual_phis): Likewise.
3033 (optimize_tail_call): Return TODOs.
3034 (execute_tail_calls): Return TODOs.
3035 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
3036 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
3038 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
3039 Cleanup cfg using TODO.
3040 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
3041 * passes.c (execute_function_todo): When cleanup did something, remove
3043 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
3044 (make_edges): Don't cleanup_cfg.
3046 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
3048 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
3049 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
3050 operate on the top of stack.
3051 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
3052 register handling of instructions that output to the second
3054 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
3055 (move_for_stack_reg): Special-case check for dead destination
3056 stack slot for constant load of 1.0 inside UNSPEC_TAN.
3058 * config/i386/i386.md (UNSPEC_TAN): New constant.
3059 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
3060 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
3061 to correctly model move of constant 1.0 to top stack slot.
3062 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
3063 (unnamed peephole2 pattern): Remove corresponding peephole2
3064 pattern that optimizes tan insn and loading of constant 1.0.
3065 (tanxf2): Use fptanxf4_i387.
3066 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
3067 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
3068 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
3069 disable patterns for SSE math.
3070 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
3071 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
3073 2007-01-18 Dirk Mueller <dmueller@suse.de>
3074 Richard Guenther <rguenther@suse.de>
3077 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
3078 * common.opt (Warray-bounds): Add new warning option.
3079 * c-opts.c (c_common_handle_option): Define -Warray-bounds
3081 * Makefile.in: make tree-vrp.o depend on toplev.h
3082 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
3084 (check_array_refs, check_array_bounds, check_array_ref): New.
3086 2007-01-18 Jan Hubicka <jh@suse.cz>
3088 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
3089 (execute_ssa_ccp): Return flags conditionally.
3090 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
3092 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
3094 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
3096 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
3097 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
3099 * Makefile.c (GTFILES): Add cfglayout.h.
3100 * gengtype.c (open_base_files): Likewise.
3101 * cfglayout.c (cfg_layout_function_footer,
3102 cfg_layout_function_header) Reindent to make gengtype happy.
3103 * cfglayout.h (cfg_layout_function_footer,
3104 cfg_layout_function_header): Add GTY(()) marker.
3106 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
3108 2007-01-18 Ben Elliston <bje@au.ibm.com>
3110 * genautomata.c (write_automata): Include xstrerror output in the
3111 error message if writing the DFA description file fails.
3113 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
3115 * config/mips/mips-protos.h (mips_output_external): Make it
3117 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
3118 * config/mips/mips.c (irix_output_external_libcall): Likewise.
3119 (extern_list): Likewise.
3120 (extern_head): Likewise.
3121 (TARGET_ASM_FILE_END): Likewise.
3122 (mips_file_end): Likewise.
3123 (mips_output_external): Rewritten.
3125 2007-01-18 Ben Elliston <bje@au.ibm.com>
3127 * genpreds.c (write_insn_preds_c): Only write out the function
3128 body for regclass_for_constraint if we have register constraints.
3130 2007-01-17 Tom Tromey <tromey@redhat.com>
3132 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
3133 * doc/install.texi (Testing): Use sourceware.org.
3134 (Binaries): Likewise.
3135 (Specific): Likewise.
3136 * doc/contrib.texi (Contributors): Use sourceware.org.
3138 2007-01-17 Anatoly Sokolov <aesok@post.ru>
3140 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
3141 (AVR_ENHANCED): Rename to ...
3142 (AVR_HAVE_MUL): ... new.
3143 (avr_enhanced_p): Rename to ...
3144 (avr_have_mul_p): ... new.
3145 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
3146 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
3147 * config/avr/avr.c (avr_enhanced_p): Rename to ...
3148 (avr_have_mul_p): ... new.
3149 (base_arch_s): Rename 'enhanced' to 'have_mul'.
3150 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
3151 'avr_enhanced_p' and 'enhanced'.
3152 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
3153 AVR_HAVE_MUL instead of AVR_ENHANCED.
3154 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3155 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
3156 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
3157 instead of AVR_ENHANCED.
3158 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
3159 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
3160 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
3162 2007-01-17 Ian Lance Taylor <iant@google.com>
3164 * vec.h (VEC_reserve_exact): Define.
3165 (vec_gc_p_reserve_exact): Declare.
3166 (vec_gc_o_reserve_exact): Declare.
3167 (vec_heap_p_reserve_exact): Declare.
3168 (vec_heap_o_reserve_exact): Declare.
3169 (VEC_OP (T,A,reserve_exact)): New static inline function, three
3171 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
3173 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
3174 (VEC_OP (T,A,copy)) [all versions]: Likewise.
3175 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
3176 * vec.c (calculate_allocation): Add exact parameter. Change all
3178 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
3179 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
3180 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
3181 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
3182 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
3183 (vec_heap_p_reserve_exact): New function.
3184 (vec_heap_o_reserve_exact): New function.
3186 2007-01-17 Jan Hubicka <jh@suse.cz>
3188 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
3190 2007-01-17 Eric Christopher <echristo@apple.com>
3192 * config.gcc: Support core2 processor.
3194 2007-01-16 Jan Hubicka <jh@suse.cz>
3196 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
3197 the names, just unlink the chain so we don't crash on dangling pointers
3200 2007-01-16 Jan Hubicka <jh@suse.cz>
3202 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
3203 * tree-pass.h: Reorder to make IPA passes appear toegher.
3204 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
3205 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
3206 parameters, do not call early inliner.
3207 * ipa-inline.c: Update comments. Include tree-flow.h
3208 (cgraph_decide_inlining): Do not compute inlining parameters.
3209 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
3210 be called with function context set up.
3211 (pass_ipa_inline): Remove unreachable functions before pass.
3212 (cgraph_early_inlining): Simplify assuming to be called from the
3214 (pass_early_inline): New pass.
3215 (cgraph_gate_ipa_early_inlining): New gate.
3216 (pass_ipa_early_inline): Turn into simple wrapper.
3217 (compute_inline_parameters): New function.
3218 (gate_inline_passes): New gate.
3219 (pass_inline_parameters): New pass.
3220 (apply_inline): Move here from tree-optimize.c
3221 (pass_apply_inline): New pass.
3222 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
3224 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
3226 (tree_function_versioning): Do not allocate dummy struct function.
3227 * tree-inline.h (optimize_inline_calls): Update prototype.
3228 * tree-optimize.c (execute_fixup_cfg): Export.
3229 (pass_fixup_cfg): Remove
3230 (tree_rest_of_compilation): Do not apply inlines.
3231 * tree-flow.h (execute_fixup_cfg): Declare.
3232 * Makefile.in (gt-passes.c): New.
3233 * passes.c: Include gt-passes.h
3234 (init_optimization_passes): New passes.
3235 (nnodes, order): New static vars.
3236 (do_per_function_toporder): New function.
3237 (execute_one_pass): Dump current pass here.
3238 (execute_ipa_pass_list): Don't dump current pass here.
3240 2007-01-16 Janis Johnson <janis187@us.ibm.com>
3242 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
3243 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
3245 2007-01-16 David Edelsohn <edelsohn@gnu.org>
3247 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
3248 (strong_alias): Define.
3249 (__gcc_qmul): Provide non-FMA for soft-float.
3255 (__gcc_qunord): New.
3265 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
3266 soft-float functions.
3267 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
3268 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
3269 about long double soft float.
3271 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
3272 Tehila Meyzels <tehila@il.ibm.com>
3274 * tree-vectorizer.h (is_pattern_stmt_p): New.
3275 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
3276 formatting (tabs instead of spaces). Cleanup and clarify setting
3277 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
3278 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
3279 (vectorizable_type_demotion): Check that types are integral.
3280 (vectorizable_type_promotion): Likewise.
3281 (vectorizable_store): Fix typo. Eliminate new-line at end of
3284 2007-01-16 Jan Hubicka <jh@suse.cz>
3286 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
3289 2007-01-15 Eric Christopher <echristo@apple.com>
3291 * ifcvt.c: Include vec.h, vecprim.h.
3292 (check_cond_move_block): New argument regs.
3293 Reorganize. Add registers used to regs.
3294 (cond_move_process_if_block): Use regs set above as
3297 2007-01-15 Eric Christopher <echristo@apple.com>
3299 * config/darwin.h: Update copyright.
3300 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
3301 (LINK_COMMAND_SPEC): Add -u.
3302 (LINK_SPEC): Fix umbrella for above.
3304 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
3306 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
3309 2007-01-15 Tom Tromey <tromey@redhat.com>
3311 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
3312 * doc/install.texi (Testing): Don't mention jacks.
3313 (Configuration): Document --enable-java-maintainer-mode. Move
3314 --with-java-home to libgcj-specific section. Document
3316 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
3318 2007-01-15 Jan Hubicka <jh@suse.cz>
3320 * tree-ssa-dce.c (DCE_TODOs): New.
3321 (propagate_necessity): Return if something changed.
3322 (eliminate_unnecessary_stmts): Likewise.
3323 (perform_tree_ssa_dce): Return TODO flags when needed.
3324 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
3326 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
3328 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
3329 (fyl2x_extend<mode>xf3_i387): New insn pattern.
3330 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
3331 patterns using X87MODEF12 mode macro. Extend operand 1
3332 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
3333 (log10<mode>2): Ditto.
3334 (log2<mode>2): Ditto.
3335 (log1p<mode>2): Ditto.
3336 (logb<mode>2): Ditto.
3337 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
3338 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
3339 (*fxtractxf3_i387): Rename from *fxtractxf3.
3340 (fxtract_extend<mode>xf3_i387): New insn pattern.
3341 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
3342 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
3343 and gen_fyl2xxf3_i387().
3345 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
3347 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
3348 * cfgloopmanip.c (fix_loop_placement): Made static. Use
3349 get_loop_exit_edges. Changed return type to bool.
3350 * cfgloop.h (fix_loop_placement): Declaration removed.
3352 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
3354 * param.h (MIN_VECT_LOOP_BOUND): New.
3355 * params.def (MIN_VECT_LOOP_BOUND): New.
3356 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
3357 argument - minimum threshold for number of iterations.
3358 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
3359 argument to declaration.
3360 * tree-vect-analyze.c (vect_analyze_operations): Check value of
3361 MIN_VECT_LOOP_BOUND.
3362 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
3363 slpeel_tree_peel_loop_to_edge with additional argument.
3364 (vect_do_peeling_for_alignment): Likewise.
3365 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
3367 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
3370 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
3373 2007-01-14 Jan Hubicka <jh@suse.cz>
3375 * tree-dfa.c (remove_referenced_var): New function.
3376 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
3377 prune referenced vars list too.
3378 * tree-flow.h (remove_referenced_var): Declare.
3380 2007-01-14 Jan Hubicka <jh@suse.cz>
3382 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
3384 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
3385 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
3386 * except.h (verify_eh_throw_table_statements): Kill prototype.
3387 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
3388 into call exrepssions.
3389 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
3391 * tree-cfg.c: Include pointer-set.h.
3392 (verify_node_sharing): Work on pointer set.
3393 (verify_eh_throw_stmt_node): New.
3394 (verify_stmts): Use pointers sets, verify throw_stmt.
3396 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
3398 * ipa-reference.c (analyze_function): Consider also addresses taken
3401 2007-01-12 Roger Sayle <roger@eyesopen.com>
3403 * c-typeck.c (null_pointer_constant_p): Replace use of
3404 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3405 (build_c_cast): Likewise.
3407 2007-01-12 Roger Sayle <roger@eyesopen.com>
3409 * tree.h (force_fit_type_double): Remove unused final argument.
3410 * c-common.c (constant_expression_warning): Replace use of
3411 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3412 (convert_and_check): Likewise.
3413 (shorten_compare): Update call to force_fit_type_double.
3414 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
3415 * convert.c (convert_to_pointer): Update call to
3416 force_fit_type_double.
3417 * fold-const.c (force_fit_type_double): Remove overflowed_const
3419 (int_const_binop, fold_convert_const_int_from_int,
3420 fold_convert_const_int_from_real, fold_div_compare,
3421 fold_sign_changed_comparison, fold_unary, fold_negate_const,
3422 fold_abs_const, fold_not_const): Remove the final argument from
3423 calls to force_fit_type_double.
3425 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
3427 * configure.ac: Set insn to "nop" for spu-*-* also.
3428 * configure: Regenerate.
3430 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
3432 * builtins.def : Add BUILT_IN_FREE.
3434 2007-01-12 Jan Hubicka <jh@suse.cz>
3436 PR tree-optimization/30443
3437 * tree-inline.c (tree_function_versioning): Do not optimize when
3438 cloning for inlining.
3440 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
3442 * doc/loop.texi: Document recording of loop exits.
3443 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
3444 (update_single_exits_after_duplication,
3445 update_single_exit_for_duplicated_loop,
3446 update_single_exit_for_duplicated_loops): Removed.
3447 (duplicate_loop_to_header_edge): Do not call
3448 update_single_exits_after_duplication and
3449 update_single_exit_for_duplicated_loops.
3450 (loop_version): Do not update single_exit information.
3451 (fix_loop_structure): Use record_loop_exits instead of
3452 mark_single_exit_loops.
3453 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
3454 the lists of loop exits.
3455 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
3456 split_edge, merge_blocks): Update the lists of loop exits.
3457 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
3458 loop_optimizer_init.
3459 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
3460 of mark_single_exit_loops.
3461 (loop_optimizer_finalize): Call release_recorded_exits.
3462 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
3463 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
3464 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
3465 update single exit information.
3466 * lambda-code.c (perfect_nestify): Ditto.
3467 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
3468 (mark_single_exit_loops): Removed.
3469 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
3470 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
3471 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
3473 (get_loop_exit_edges, single_exit): Use recorded exit lists.
3474 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
3475 (verify_loop_structure): Verify consistency of the exit lists.
3476 (flow_loops_find): Use alloc_loop. Initialize exits hash.
3477 (set_single_exit): Removed.
3478 * cfgloop.h (struct loop_exit): New function.
3479 (struct loop): single_exit_ field replaced by exits field.
3480 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
3481 (struct loops): Added exits hash.
3482 (mark_single_exit_loops, set_single_exit): Declaration removed.
3483 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
3485 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3487 * doc/invoke.texi: Avoid use of @headitem.
3489 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3491 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
3493 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3495 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
3498 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3499 Richard Sandiford <richard@codesourcery.com>
3501 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
3502 * config/m68k/m68k.opt: Resort options.
3503 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
3505 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3506 Richard Sandiford <richard@codesourcery.com>
3508 * doc/invoke.texi: Document the macros that are defined by
3509 m68k's -mtune and -mhard-float options.
3510 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
3511 (m68k_cpp_cpu_family): Likewise.
3512 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
3513 of __ucfv*__ macros. Define __mcffpu__ if generating code for
3514 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
3515 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
3516 (m68k_cpp_cpu_family): Likewise.
3518 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3520 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
3521 macros besides mc68000 as tuning macros. Use a switch statement
3522 to set them and mcpu32.
3524 2007-01-12 Julian Brown <julian@codesourcery.com>
3526 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
3527 (TARGET_68040_ONLY): Rename to...
3528 (TARGET_68040): ...this.
3529 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
3530 * config/m68k/m68k.md: Likewise.
3532 2007-01-12 Julian Brown <julian@codesourcery.com>
3533 Nathan Sidwell <nathan@codesourcery.com>
3534 Richard Sandiford <richard@codesourcery.com>
3536 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
3537 the -mcpu= argument associated with the --with-cpu setting.
3538 Define M68K_DEFAULT_TUNE to the default -mtune= option,
3539 if different from the one implied by the -mcpu setting.
3540 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
3541 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
3542 * doc/invoke.texi: Mention ColdFire in the introduction to the
3543 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
3544 -mno-div and -mhard-float options. Make -m68881 a synonym for
3545 -mhard-float. Document the previously-undocumented -m5206e,
3546 -m528x, -m5307 and -m5407 options. Tweak the existing option
3547 documentation for consistency.
3548 * doc/install.texi: Mention new --with-cpu arguments.
3549 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
3550 default CPU if neither -mcpu nor -march are specified.
3551 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
3552 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
3553 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
3554 using m68k_tune to decide between families that implement the
3555 same ISA. Use m68k_tune to set __mcfv4e__.
3556 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
3557 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
3558 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
3559 (FL_ISA_C, FL_ISA_MMU): New macros.
3560 (MASK_COLDFIRE): Delete.
3561 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
3562 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
3563 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
3564 (TARGET_HARD_FLOAT): Do not define here.
3565 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
3566 (TUNE_68000): New macro.
3567 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
3568 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
3569 (TUNE_CFV2): Redefine in terms of m68k_tune.
3570 (uarch_type, target_device, fpu_type): New enums.
3571 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
3572 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
3573 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
3574 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
3575 (FL_FOR_isa_c): New macros.
3576 (m68k_isa): New enum.
3577 (m68k_target_selection): New structure.
3578 (all_devices, all_isas, all_microarchs): New tables.
3579 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
3580 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
3581 (MASK_ALL_CPU_BITS): Delete.
3582 (m68k_find_selection): New function.
3583 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
3584 Map the legacy target options to a combination of the new ones.
3585 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
3586 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
3587 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
3589 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
3590 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
3592 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
3593 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
3594 * config/m68k/m68k-devices.def: New file.
3596 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3597 Nathan Sidwell <nathan@codesourcery.com>
3599 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
3600 (SUBTARGET_EXTRA_SPECS): New macros.
3601 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
3602 use %(asm_cpu_spec) instead.
3603 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
3604 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
3605 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
3606 (EXTRA_SPECS): Rename to...
3607 (SUBTARGET_EXTRA_SPECS): ...this.
3609 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
3610 Richard Sandiford <richard@codesourcery.com>
3611 Julian Brown <julian@codesourcery.com>
3613 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
3614 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
3615 to the configuration's default CPU.
3616 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
3617 Remove default masks.
3618 (m680[012]0-*-*): Set the default with_cpu to the first part of
3620 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
3621 (m68k*-*-linux): Extend the --with-cpu handling to...
3622 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
3623 and m68010. Don't set target_cpu_default2.
3624 * doc/install.texi: Document --with-cpu for m68k.
3625 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
3626 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
3627 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
3628 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
3629 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
3630 (ASM_SPEC): Remove use of %(asm_cpu_default).
3631 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
3632 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
3633 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
3634 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
3635 __HAVE_FPU__ if TARGET_HARD_FLOAT.
3636 (TARGET_DEFAULT): Delete.
3637 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
3638 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
3639 (CPP_CPU_SPEC): Delete.
3640 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
3641 the appropriate string.
3642 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
3643 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
3644 (ASM_SPEC): Don't use %(asm_default_spec).
3645 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
3646 TARGET_DEFAULT and add MASK_68881.
3647 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
3649 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3651 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
3652 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
3653 MASK_68010 alongside MASK_68020.
3654 * doc/invoke.texi: Document -m68010.
3655 * config/m68k/m68k.opt (m68010): New.
3656 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
3658 (TUNE_68010): New macro.
3659 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
3660 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
3661 (M68K_CPU_m68332): Add MASK_68010.
3662 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
3663 fallback definition.
3664 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
3666 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
3667 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
3668 to all entries that use MASK_68020.
3669 (output_move_simode_const, output_move_himode, output_move_qimode)
3670 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
3671 instead of TARGET_68020 to select clr behavior. Remove comment
3672 about there being no TARGET_68010.
3673 * config/m68k/m68k.md: Likewise throughout.
3675 2007-01-12 Julian Brown <julian@codesourcery.com>
3677 * config/m68k/m68k.h (TARGET_ISAB): New macro.
3678 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
3679 * config/m68k/m68k.md: Likewise.
3681 2007-01-12 Julian Brown <julian@codesourcery.com>
3683 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
3684 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
3686 2007-01-12 Julian Brown <julian@codesourcery.com>
3688 * config/m68k/m68k.h (TUNE_68040_60): New macro.
3689 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
3690 * config/m68k/m68k.md: Likewise.
3692 2007-01-12 Julian Brown <julian@codesourcery.com>
3693 Richard Sandiford <richard@codesourcery.com>
3695 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
3696 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
3697 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
3699 (TARGET_CPU32): Rename to...
3700 (TUNE_CPU32): ...this.
3701 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
3702 (TUNE_CFV2): New macros.
3703 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
3704 remove conditions that are implied by TARGET_68020.
3705 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
3706 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
3707 (m68k_output_function_epilogue): Likewise.
3708 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
3709 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
3710 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
3711 non-68000 timings. Refactor multiplication and division costs.
3712 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
3713 TUNE_CPU32 instead of TARGET_CPU32.
3714 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
3715 and TUNE_68060 instead of TARGET_68060.
3716 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
3717 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
3719 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
3720 && !TARGET_COLDFIRE" to choose between moveq and clr.
3721 Likewise in the unnamed movsf pattern.
3722 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
3723 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
3726 2007-01-12 Richard Sandiford <richard@codesourcery.com>
3728 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
3729 of tabbing before backslashes.
3731 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3733 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
3734 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
3735 (ASM_OUTPUT_INTERNAL_LABEL): Define.
3737 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3739 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
3740 find_interesting_uses_cond.
3741 (find_interesting_uses_cond): Use extract_cond_operands.
3742 (rewrite_use_compare): Use extract_cond_operands and
3743 force_gimple_operand_bsi. Do not call update_stmt.
3744 (determine_use_iv_cost_condition): Use extract_cond_operands.
3745 Return cheaper of using original bound and changing the exit bound.
3747 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
3749 PR tree-optimization/29516
3750 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
3751 most_expensive_mult_to_index, addr_to_parts,
3752 create_mem_ref, maybe_fold_tmr): Make the type of
3753 fields of TARGET_MEM_REF sizetype.
3754 (move_fixed_address_to_symbol, move_pointer_to_base):
3756 * tree.def (TARGET_MEM_REF): Add comment on types of
3759 2007-01-11 Joseph Myers <joseph@codesourcery.com>
3761 * c-common.c (vector_types_convertible_p): Treat opaque types as
3762 always convertible if they have the same size, but not otherwise.
3764 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
3766 * ifcvt.c (struct noce_if_info): Add comments to the fields.
3767 Remove the b_unconditional field.
3768 (noce_try_sign_mask): Do not look at b_unconditional.
3769 (noce_process_if_block): Do not use merge_if_blocks. Update
3770 the CFG here. Do not set b_unconditional.
3771 (cond_move_process_if_block): Likewise.
3772 (find_cond_trap): Likewise.
3773 (check_cond_move_block): Require simple jump insns at the end
3776 2007-01-11 Jan Hubicka <jh@suse.cz>
3778 PR tree-optimization/1046
3779 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
3780 when alias info is not ready.
3781 (pass_tail_recursion): Do not require aliasing.
3782 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
3783 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
3784 * tree-ssa-copy.c (pass_copy_prop): Likewise.
3785 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
3786 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
3787 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
3788 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
3790 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
3791 When aliasing is not build, mark statement as volatile.
3793 2007-01-11 Tom Tromey <tromey@redhat.com>
3795 PR preprocessor/15185, PR preprocessor/20989:
3796 * doc/cppopts.texi <-MT>: Update description of algorithm for
3797 computing default target.
3798 <-M, -MD>: Reword "basename" text.
3800 2007-01-11 Roger Sayle <roger@eyesopen.com>
3802 * builtins.c (expand_builtin_pow, expand_builtin_powi,
3803 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
3804 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
3805 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
3806 fold_builtin_bswap, real_constp, fold_builtin_pow,
3807 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
3808 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
3809 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
3810 * convert.c (convert_to_pointer): Likewise.
3811 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
3812 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
3813 const_binop, fold_convert_const_int_from_int,
3814 fold_convert_const_int_from_real,
3815 fold_convert_const_real_from_real, sign_bit_p,
3816 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
3817 fold_sign_changed_comparison, fold_unary, fold_comparison,
3818 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
3819 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
3820 * print-tree.c (print_node_brief, print_node): Likewise.
3821 * stor-layout.c (place_field, layout_type): Likewise.
3822 * tree-chrec.c (keep_cast): Likewise.
3823 * tree.c (build_vector, build_real, build_real_from_int_cst,
3824 build_complex): Likewise.
3826 2007-01-11 Roger Sayle <roger@eyesopen.com>
3828 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
3829 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
3831 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
3833 * configure.ac (strict1_warn): Rename to strict_warn.
3834 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
3835 cc_set_by_configure, quoted_cc_set_by_configure,
3836 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
3837 all_boot_languages, all_stagestuff): Remove.
3838 (target_list): Remove bootstrap targets.
3839 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
3840 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
3841 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
3842 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
3843 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
3844 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
3845 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
3846 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
3847 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
3848 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
3849 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
3850 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
3851 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
3852 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
3853 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
3854 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
3855 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
3856 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
3857 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
3858 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
3859 (mostlyclean): Adjust.
3860 (clean, distclean): Don't mention bootstrap stuff.
3861 * configure: Regenerate.
3862 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
3863 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
3864 treelang/config-lang.in (stagestuff): Remove.
3865 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
3866 stageprofile, stagefeedback, stagestuff): Remove mention.
3868 2007-01-11 Nick Clifton <nickc@redhat.com>
3870 * config/mcore/predicates.md (mcore_general_movesrc_operand):
3872 (mcore_general_movdst_operand): Do not accept CONST_INTs.
3873 (mcore_arith_K_S_operand): Run the test for the S constraint not
3874 the test for the M constraint.
3875 (mcore_addsub_operand): Do not accept integer values that are
3876 larger than 32 bits.
3877 * config/mcore/mcore.md: Remove unused constraints from split.
3878 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
3880 (allocate_stack): Likewise.
3881 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
3882 operands to 32 bits.
3883 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
3885 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
3886 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
3887 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
3888 mcore_output_bclri, mcore_output_andn, output_inline_const,
3889 mcore_output_move, mcore_output_movedouble): Likewise.
3890 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
3891 (output_inline_const): Likewise.
3892 (output_inline_const): Fix format strings used in sprintf
3894 * config/mcore/mcore-protos.h: Update prototypes for changed
3895 functions in mcore.c.
3896 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
3897 HOST_WIDE_INT and not int.
3898 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
3899 CONST_OK_FOR_N): Likewise.
3900 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
3901 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
3904 2007-01-10 Jan Hubicka <jh@suse.cz>
3906 * tree-vrp.c (remove_range_assertions): Release defs.
3907 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
3908 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
3910 2007-01-10 Paul Brook <paul@codesourcery.com>
3912 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
3913 * config/arm/arm.md (mulsi3subsi): New insn.
3915 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
3917 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
3918 (tree_transform_and_unroll_loop): New.
3919 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
3922 2007-01-10 Robert Kennedy <jimbob@google.com>
3924 * fold-const.c (fold_comparison): Fold comparisons like (x *
3925 1000 < 0) to (x < 0).
3927 2007-01-10 Ian Lance Taylor <iant@google.com>
3929 * tree-pretty-print.c (dump_generic_node): Print parentheses when
3930 operands have the same priority.
3932 2007-01-10 Tom Tromey <tromey@redhat.com>
3934 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
3935 (fold_binary): Likewise.
3936 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
3937 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
3938 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
3939 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
3940 * langhooks.h (struct lang_hooks): Removed field
3941 'can_use_bit_fields_p'.
3943 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
3945 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
3947 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
3949 * function.c (get_last_funcdef_no): New function.
3950 * function.h (get_last_funcdef_no): Declare.
3951 * tree-inline.c (initialize_cfun): Add initialization.
3952 (tree_function_versioning): Cleanup.
3954 2007-01-10 Jan Hubicka <jh@suse.cz>
3956 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
3959 2007-01-10 Sa Liu <saliu@de.ibm.com>
3960 Ben Elliston <bje@au.ibm.com>
3962 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
3963 for SAVE_NONLOCAL and Pmode for any other save level.
3964 * spu-protos.h (spu_restore_stack_block): Declare.
3965 * spu.md (save_stack_block): Remove.
3966 (restore_stack_block): Call spu_restore_stack_block.
3967 * spu.c (spu_restore_stack_block): New function.
3968 (spu_expand_epilogue): Remove old comment.
3970 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
3972 PR tree-optimization/30322
3973 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
3974 (cand_value_at): Return the value as aff_tree.
3975 (may_eliminate_iv): Convert the bound from aff_tree to tree.
3976 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
3977 aff_combination_mult): New functions.
3978 (aff_combination_add): Use aff_combination_add_cst.
3979 (aff_combination_convert): Allow conversions to a wider type.
3980 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
3981 * tree-affine.h (aff_combination_mult): Declare.
3983 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
3985 * doc/tm.texi: Update documentation to reflect reality of exec
3986 and start file search behaviours. Update copyright year.
3987 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
3988 header file directories.
3990 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
3992 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
3993 (*cosxf2): Rename to cosxf2_i387.
3994 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
3995 using X87MODEF12 mode macro. Rename patterns to
3996 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
3998 (*cosdf2, *cossf2): Ditto.
3999 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
4000 to match extended input operands.
4001 (sincos<mode>3): New expander.
4002 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
4003 insn patterns and corresponding splitters.
4005 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
4007 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
4008 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
4010 2007-01-09 Nicolas Pitre <nico@cam.org>
4013 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
4015 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4017 * target.h (struct gcc_target): New field library_rtti_comdat.
4018 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
4019 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
4020 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
4021 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4023 2007-01-08 Geoffrey Keating <geoffk@apple.com>
4025 * doc/invoke.texi (Optimize Options): Correct description of -O0.
4027 2007-01-08 Richard Guenther <rguenther@suse.de>
4029 * tree.h (force_fit_type_double): Export.
4030 (force_fit_type): Remove.
4031 * fold-const.c (force_fit_type_double): New function.
4032 (force_fit_type): Remove.
4033 (int_const_binop): Use it.
4034 (fold_convert_const_int_from_int): Likewise.
4035 (fold_convert_const_int_from_real): Likewise.
4036 (fold_div_compare): Likewise.
4037 (fold_sign_changed_comparison): Likewise.
4038 (fold_unary): Likewise.
4039 (fold_negate_const): Likewise.
4040 (fold_abs_const): Likewise.
4041 (fold_not_const): Likewise.
4042 * c-common.c (shorten_compare): Use force_fit_type_double.
4043 * convert.c (convert_to_pointer): Likewise.
4045 2007-01-08 Richard Guenther <rguenther@suse.de>
4047 * tree.h (build_int_cst_wide_type): Export.
4048 * tree.c (build_int_cst_wide_type): New function.
4049 (build_int_cst_wide): Fix comment.
4050 * builtins.c (fold_builtin_object_size): Use build_int_cst
4051 to build -1 or 0 of the correct type. Use fit_double_type
4052 to check for overflow.
4053 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
4055 (decode_field_reference): Likewise.
4056 (all_ones_mask_p): Likewise.
4057 (native_interpret_int): Use build_int_cst_wide_type.
4058 (fold_binary): Use build_int_cst_type to build an all-ones
4060 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
4062 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
4064 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
4066 2007-01-09 Ben Elliston <bje@au.ibm.com>
4068 * genautomata.c (STATS_OPTION): New option.
4069 (stats_flag): New flag.
4070 (gen_automata_option): Handle it.
4071 (initiate_automaton_gen): Ditto.
4072 (write_automata): Output statistics only if stats_flag is
4073 set. Likewise, output time statistics only if time_flag is set.
4074 * doc/md.texi (Processor pipeline description): Document new flag.
4076 2007-01-08 Richard Guenther <rguenther@suse.de>
4078 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
4079 * tree.c (build_int_cst_type): Likewise.
4080 (size_in_bytes): Don't call force_fit_type on the result.
4081 (int_fits_type_p): Use fit_double_type.
4082 * fold-const.c (fit_double_type): New function.
4083 (force_fit_type): Use it.
4084 * tree.h (fit_double_type): Export.
4086 2007-01-08 Jan Hubicka <jh@suse.cz>
4088 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
4089 * ipa.c (function_and_variable_visibility): Fix return type.
4091 2007-01-08 Richard Guenther <rguenther@suse.de>
4093 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
4094 of offset to build the index.
4095 * tree-pretty-print.c (dump_generic_node): Don't build negated
4096 const just for printing.
4097 * c-pretty-print.c (pp_c_integer_constant): Likewise.
4098 * builtins.c (fold_builtin_int_roundingfn): Check if result
4099 fits the type by using force_fit_type and comparing the result.
4100 * predict.c (predict_loops): Use compare_tree_int for comparison.
4101 * tree.c (build_int_cst): Fall back to integer_type_node for
4103 (build_int_cst_wide): Assert type is non-null.
4105 2007-01-08 Roberto Costa <roberto.costa@st.com>
4107 * tree-vrp.c (extract_range_from_cond_expr): New.
4108 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
4109 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
4111 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
4112 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
4113 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
4114 used as expressions.
4115 * tree-object-size.c (cond_expr_object_size): New.
4116 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
4118 2007-01-08 Jan Hubicka <jh@suse.cz>
4120 * tree-ssa-forwprop.c (forward_propagate_into_cond,
4121 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
4124 2007-01-08 Richard Guenther <rguenther@suse.de>
4126 PR tree-optimization/23603
4127 * tree-vrp.c (set_value_range_to_truthvalue): New function.
4128 (extract_range_from_binary): Fall back to truthvalue instead of
4129 varying for TRUTH_*_EXPR.
4130 (extract_range_from_comparison): Fall back to truthvalue instead of
4132 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
4133 if all visited PHI values were constant.
4135 2007-01-08 Jan Hubicka <jh@suse.cz>
4137 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
4138 (cgraph_analyze_function): break out from ...
4139 (cgraph_finalize_compilation_unit): ... here.
4140 (cgraph_expand_function): Remove forgoten commented out line.
4141 (cgraph_optimize): Analyze functions.
4143 2007-01-08 Jan Hubicka <jh@suse.cz>
4145 * tree-pas.h (TODO_remove_function): New flag.
4146 (TODO_update*): Renumber.
4147 (pass_ipa_increase_alignment,
4148 pass_ipa_function_and_variable_visibility): New passes.
4149 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
4150 (cgraph_function_and_variable_visibility): Move to ipa.c
4151 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
4152 cgraph_increase_alignment.
4153 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
4154 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
4155 calling tree-inline.
4156 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
4157 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
4158 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
4159 (gate_increase_alignment): New function.
4160 (pass_ipa_increase_alignment): New pass.
4161 * ipa.c: Inline tree-pass.h and timevar.h
4162 (function_and_variable_visibility): Move here from cgraphunit.c
4163 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
4164 * passes.c (init_optimization_passes): Add the two new passes.
4165 (execute_todo): Handle cgraph_remove_functions.
4167 2007-01-08 Nick Clifton <nickc@redhat.com>
4169 * config/frv/predicates.md (reg_or_0_operand): Accept
4172 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
4174 * config/bfin/rtems.h, config/bfin/t-rtems: New.
4175 * config.gcc: Add bfin*-rtems*.
4177 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4179 * c.opt: Add -flax-vector-conversions.
4180 * c-typeck.c (convert_for_assignment): Pass flag to
4181 vector_types_convertible_p to allow emission of note.
4182 (digest_init): Likewise.
4183 * c-opts.c: Handle -flax-vector-conversions.
4184 * c-common.c (flag_lax_vector_conversions): New.
4185 (vector_types_convertible_p): Unless -flax-vector conversions
4186 has been passed, disallow conversions between vectors with
4187 differing numbers of subparts and/or element types. If such
4188 a conversion is disallowed, possibly emit a note on the first
4189 occasion only to inform the user of -flax-vector-conversions.
4190 The new last argument specifies this.
4191 * c-common.h (flag_lax_vector_conversions): New.
4192 (vector_types_convertible_p): Add extra argument.
4193 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
4194 char_type_node for V*QI type vectors.
4195 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
4196 Update to satisfy new typechecking rules.
4197 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
4199 * doc/invoke.texi (C Dialect Options): Document
4200 -flax-vector-conversions.
4202 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
4204 PR tree-optimization/29877
4205 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
4206 a register variable on the RHS to not be replaceable.
4208 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
4209 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
4210 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
4211 const_simm15, const_pow2, const_npow2): Added.
4212 * config/score/misc.md (insv, extv, extzv, movmemsi,
4213 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
4214 * config/score/score.c (score_address_cost, score_select_cc_mode):
4216 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
4217 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
4218 Update score_print_operand.
4219 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
4220 Adjust register allocate order and update some macro define.
4221 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
4222 mdx_block_move_straight, mdx_block_move_loop_head,
4223 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
4224 mdx_block_move): Added.
4225 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
4226 mdp_limm): Updated and fix some bug and typo.
4227 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
4228 (movsf, movdf, doloop_end): Added.
4230 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
4232 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
4233 config/arm/thumb2.md: Fix comment typos.
4234 * doc/extend.texi: Fix a typo.
4236 2007-01-07 Eric Christopher <echristo@apple.com>
4238 * configure.ac: Check for __stack_chk_fail for darwin.
4239 * configure: Regenerate.
4241 2007-01-07 Richard Guenther <rguenther@suse.de>
4243 * tree-vrp.c (extract_range_from_assert): CSE calls to
4244 compare_values where possible.
4245 (extract_range_from_unary_expr): Likewise.
4247 2007-01-07 Anatoly Sokolov <aesok@post.ru>
4249 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
4250 * config/avr/avr.c (call_insn_operand): Delete function.
4251 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
4252 instead of register number. Use predicates.
4253 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
4254 call_insn_operand): Add.
4256 2007-01-06 Jan Hubicka <jh@suse.cz>
4258 * tree-pass.h (pass_build_cgraph_edges): Declare.
4259 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
4260 (visited_nodes): Remove.
4261 (cgraph_create_edges): Move to cgraphbuild.c; rename to
4262 build_cgrpah_edges; make visited_nodes local.
4263 (cgraph_process_new_functions): DO not call initialize_inline_failed.
4264 (record_references_in_initializer): Move to cgraphbuild.c
4265 (initialize_inline_failed, rebuild_cgraph_edges,
4266 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
4267 (verify_cgraph_node): Make visited_nodes local.
4268 (cgraph_analyze_function): Do not call cgraph_create_edges and
4269 initialize_inline_failed.
4270 (cgraph_expand_function): Do not call cgraph_lower_function;
4271 assert that function is already lowered.
4272 * Makefile.in (cgraphbuild.o): New.
4273 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
4274 at the end of lowering passes.
4276 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
4278 * ifcvt.c (cond_move_convert_if_block): New function, code
4279 factored out from...
4280 (cond_move_process_if_block): ...here. Call the new function
4281 on the THEN and ELSE blocks.
4282 (merge_if_block): Do not copy global_live_at_end, merge_blocks
4283 already takes care of this.
4285 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4288 * tree.h (TREE_OVERFLOW_P): New.
4289 * c-typeck.c (parser_build_unary_op): Warn only if result
4290 overflowed and operands did not.
4291 (parser_build_binary_op): Likewise.
4292 (convert_for_assignment): Remove redundant overflow_warning.
4293 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
4295 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4297 * c-typeck.c (store_init_value): Split over two lines to follow
4298 the GNU coding style.
4300 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
4302 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
4303 __GXX_EXPERIMENTAL_CXX0X__.
4304 * doc/cpp.texi: Same.
4306 2007-01-05 Richard Guenther <rguenther@suse.de>
4309 * tree.c (get_narrower): Do not construct COMPONENT_REFs
4310 with mismatched types. Instead explicitly build a
4311 conversion NOP_EXPR.
4313 2007-01-05 Ian Lance Taylor <iant@google.com>
4315 * c-common.c (decl_with_nonnull_addr_p): New function.
4316 (c_common_truthvalue_conversion): Call it.
4317 * c-typeck.c (build_binary_op): Likewise.
4318 * c-common.h (decl_with_nonnull_addr_p): Declare.
4320 2007-01-05 Jakub Jelinek <jakub@redhat.com>
4323 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
4326 2007-01-05 Joel Brobecker <brobecker@adacore.com>
4328 * doc/install.texi (Final install): Document the fact that
4329 the GNAT runtime should not be stripped.
4331 2007-01-04 Jan Hubicka <jh@suse.cz>
4333 * tree-inline.c (fold_marked_statements): Update operand caches
4334 and EH after folding
4336 2007-01-04 Ian Lance Taylor <iant@google.com>
4338 * c-common.c (check_function_nonnull): Whitespace fix.
4340 2007-01-04 Jan Hubicka <jh@suse.cz>
4342 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
4343 comitted older version of patch.
4344 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
4345 function, celanup cfg and collect garbage.
4347 2007-01-04 Mike Stump <mrs@apple.com>
4349 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
4352 2007-01-04 Eric Christopher <echristo@apple.com>
4354 * libgcc2.c (__bswapsi2): Use SItype.
4355 (__bswapdi2): Use DItype.
4356 * libgcc2.h: Update for above.
4358 2007-01-04 Paul Brook <paul@codesourcery.com>
4360 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
4361 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
4362 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
4363 umulsidi3, umulsidi3adddi, smulsi3_highpart,
4364 umulsi3_highpart): Make conditional on !arm_arch6.
4365 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
4366 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
4367 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
4368 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
4369 umulsi3_highpart_v6): New insns.
4371 2007-01-04 Roger Sayle <roger@eyesopen.com>
4373 * fold-const.c (fold_convert): When casting an expression to void,
4374 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
4375 have a type. Instead of attempting to build a NOP_EXPR, return
4376 these "special" trees directly.
4378 2007-01-04 Joseph Myers <joseph@codesourcery.com>
4380 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
4381 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
4382 rs6000_cost->fp not 0 as adjustment for outer NEG.
4384 2007-01-04 Jan Hubicka <jh@suse.cz>
4386 * cgraph.c (cgraph_release_function_body): New function.
4387 (cgraph_remove_node): Use it.
4388 * cgraph.h (cgraph_release_function_body): Declare.
4389 * cgraphunit.c (cgraph_expand_function): Use it.
4390 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
4391 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
4392 is initialized and while compilation of other function is running.
4393 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
4394 statement CFG annotations from here to ...
4395 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
4397 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
4399 * cfgloop.h (enum li_flags): Make the constants powers of two.
4401 2007-01-04 Jan Hubicka <jh@suse.cz>
4403 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
4405 (fold_marked_statements): New function.
4406 (optimize_inline_calls, tree_function_versioning): Fold new statements.
4407 * tree-inline.h (copy_body_data): Add statements_to_fold.
4409 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
4411 * config.gcc: Mention libgcc/config.host.
4412 * Makefile.in: Update comments mentioning libgcc.
4413 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
4414 (all.cross, start.encap, rest.encap, rest.cross): Update
4415 dependencies for libgcc move.
4416 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
4417 (clean-target, clean-target-libgcc): Delete.
4418 (srcdirify, GCC_EXTRA_PARTS): New macros.
4419 (libgcc-support, libgcc.mvars): New rules.
4420 (distclean): Remove mention of mklibgcc.
4421 (install): Don't reference INSTALL_LIBGCC.
4422 (install-common): Don't reference EXTRA_PARTS.
4423 (install-libgcc, install-multilib): Delete rules.
4424 * mklibgcc.in: Delete file.
4425 * doc/configfiles.texi: Don't mention mklibgcc.
4427 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
4428 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
4429 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
4430 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
4431 (mklibgcc, ldblspecs): Likewise.
4433 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
4434 (posixpre.def): Use $(T).
4435 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
4436 (SHLIB_SRC, SHLIB_INSTALL): Delete.
4437 (SHLIB_LINK): Make dummy.
4438 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
4440 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
4442 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
4443 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
4444 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
4445 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
4447 * configure.ac (all_outputs): Remove mklibgcc.
4448 * configure: Regenerated.
4450 2007-01-03 Josh Conner <jconner@apple.com>
4453 * calls.c (compute_argument_addresses): Set stack and stack_slot
4454 for partial args, too.
4455 (store_one_arg): Use locate.size.constant for the size when
4456 generating a save_area.
4458 2007-01-03 Robert Kennedy <jimbob@google.com>
4460 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
4461 whose definitions are deleted due to basic block merging.
4463 2007-01-03 Paul Brook <paul@codesourcery.com>
4466 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
4468 (use_return_insn): Return 0 for Thumb interrupt functions.
4469 (print_multi_reg): Add rfe argument for IRQ returns.
4470 (arm_output_epilogue): Pop interrupt return address directly into PC.
4471 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
4473 2007-01-03 Paul Brook <paul@codesourcery.com>
4475 Merge from sourcerygxx-4_1.
4476 * config/arm/thumb2.md: New file.
4477 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
4479 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
4480 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4481 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
4482 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
4484 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
4485 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
4487 * config/arm/ieee754-sf.S: Ditto.
4488 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
4489 (thumb1_base_register_rtx_p): ... to this.
4490 (thumb_index_register_rtx_p): Rename...
4491 (thumb1_index_register_rtx_p): ... to this.
4492 (thumb_output_function_prologue): Rename...
4493 (thumb1_output_function_prologue): ... to this.
4494 (thumb_legitimate_address_p): Rename...
4495 (thumb1_legitimate_address_p): ... to this.
4496 (thumb_rtx_costs): Rename...
4497 (thumb1_rtx_costs): ... to this.
4498 (thumb_compute_save_reg_mask): Rename...
4499 (thumb1_compute_save_reg_mask): ... to this.
4500 (thumb_final_prescan_insn): Rename...
4501 (thumb1_final_prescan_insn): ... to this.
4502 (thumb_expand_epilogue): Rename...
4503 (thumb1_expand_epilogue): ... to this.
4504 (arm_unwind_emit_stm): Rename...
4505 (arm_unwind_emit_sequence): ... to this.
4506 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
4507 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
4508 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
4509 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
4510 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
4511 thumb2_output_casesi): New functions.
4512 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
4513 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
4514 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
4515 THUMB2_WORK_REGS): Define.
4516 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
4517 arm_condexec_mask, arm_condexec_masklen)): New variables.
4518 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
4519 (arm_override_options): Check new CPU capabilities.
4520 Set new architecture flag variables.
4521 (arm_isr_value): Handle v7m interrupt functions.
4522 (user_return_insn): Return 0 for v7m interrupt functions. Handle
4524 (const_ok_for_arm): Handle Thumb-2 constants.
4525 (arm_gen_constant): Ditto. Use movw when available.
4526 (arm_function_ok_for_sibcall): Return false for v7m interrupt
4528 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
4529 (thumb_find_work_register, arm_load_pic_register,
4530 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
4531 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
4532 print_multi_reg, output_mov_long_double_fpa_from_arm,
4533 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
4534 output_mov_double_fpa_from_arm, output_move_double,
4535 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
4536 output_return_instruction, arm_output_function_prologue,
4537 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
4538 arm_output_mi_thunk, thumb_set_return_address): Ditto.
4539 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
4540 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
4541 (arithmetic_instr, shift_op): Use arm_shift_nmem.
4542 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
4544 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
4545 (thumb_expand_prologue): Use thumb_set_frame_pointer.
4546 (arm_file_start): Output directive for unified syntax.
4547 (arm_unwind_emit_set): Handle stack alignment instruction.
4548 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
4549 Add v6t2, v7, v7a, v7r and v7m.
4550 (RETLDM): Add Thumb-2 code.
4551 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
4552 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
4553 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
4554 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
4555 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
4556 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
4557 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
4558 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
4559 ADJUST_INSN_LENGTH): Define.
4560 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
4561 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
4562 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
4563 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
4564 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
4565 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
4566 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
4567 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
4568 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
4569 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
4570 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
4571 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
4572 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
4573 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
4574 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
4575 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
4576 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
4578 * config/arm/arm-tune.md: Regenerate.
4579 * config/arm/arm-protos.h: Update prototypes.
4580 * config/arm/vfp.md: Enable patterns for Thumb-2.
4581 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
4582 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
4583 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
4584 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
4585 * config/arm/libunwind.S: Add Thumb-2 code.
4586 * config/arm/constraints.md: Update include Thumb-2.
4587 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
4588 * config/arm/ieee754-df.S: Ditto.
4589 * config/arm/bpabi.S: Ditto.
4590 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
4591 * config/arm/predicates.md (low_register_operand,
4592 low_reg_or_int_operand, thumb_16bit_operator): New.
4593 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
4594 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
4595 * config/arm/t-arm-elf: Add armv7 multilib.
4596 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
4598 (UNSPEC_STACK_ALIGN, ce_count): New.
4599 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
4600 arm_zero_extendsidi2, arm_zero_extendqidi2): New
4602 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
4603 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
4604 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
4605 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
4606 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
4607 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
4609 * doc/extend.texi: Document ARMv7-M interrupt functions.
4610 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
4612 2007-01-03 Jakub Jelinek <jakub@redhat.com>
4614 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
4615 (struct _Unwind_Context): Rename args_size to flags, remove
4616 signal_frame field, add a new args_size field and version field.
4617 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
4618 _Unwind_IsExtendedContext): New inline functions.
4619 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
4620 Assume by_value array is only present if _Unwind_IsExtendedContext.
4621 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
4622 _Unwind_IsSignalFrame.
4623 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
4624 (uw_update_context_1): Use _Unwind_SetSignalFrame.
4625 (uw_init_context_1): Initialize context->flags to
4626 EXTENDED_CONTEXT_BIT.
4627 * config/rs6000/linux-unwind.h (frob_update_context): Use
4628 _Unwind_SetSignalFrame.
4630 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
4633 * gimplify.c (gimplify_modify_expr_complex_part): Move below
4634 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
4637 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
4639 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
4640 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
4641 RET, AT_SP): Remove.
4642 * config/i386/i386.md (*sse_prologue_save_insn): Use return
4645 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4646 * cfgbuild.c (find_basic_blocks): Likewise.
4647 * cfgrtl.c (rtl_create_basic_block): Likewise.
4648 * function.c (temp_slots_at_level): Likewise.
4649 * reg-stack.c (stack_regs_mentioned): Likewise.
4650 * regclass.c (allocate_reg_info): Likewise.
4651 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4652 set_bb_for_stmt, move_block_to_fn): Likewise.
4653 * tree-complex.c (tree_lower_complex): Likewise.
4654 * vec.h (VEC_safe_grow_cleared): New.
4656 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
4658 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
4660 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
4661 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
4663 * cfgloopmanip.c (scale_loop_frequencies): Export.
4664 (loopify): Scale the frequencies by prescribed coefficients.
4665 (set_zero_probability): New function.
4666 (duplicate_loop_to_header_edge): Improve updating of frequencies.
4667 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
4668 and frequencies according to arguments.
4669 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
4671 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
4672 * modulo-sched.c (sms_schedule): Set probabilities for entering
4673 versioned loop correctly.
4674 * tree-vect-transform.c (vect_transform_loop): Ditto.
4675 * cfgloop.h (loopify, loop_version): Declaration changed.
4676 (scale_loop_frequencies): Declared.
4678 2007-01-02 Jan Hubicka <jh@suse.cz>
4680 * cgraph.c: Include tree-flow.h
4681 (cgraph_add_new-function): Handle IPA_SSA mode; execute
4683 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
4684 * tree-pass.h (pass_all_early_optimizations): Declare.
4685 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
4687 (cgraph_analyze_function): Do early_local_passes.
4688 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4689 Do not add referenced vars.
4690 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
4692 (gate_all_early_local_passes): New gate.
4693 (pass_early_local_passes): Use new gate.
4694 (execute_early_local_optimizations): New functions.
4695 (gate_all_early_optimizations): New gate.
4696 (pass_all_early_optimizations): New pass.
4697 (execute_free_datastructures): Free SSA only when initialized.
4698 (gate_init_datastructures): Init only when optimizing.
4699 (tree_lowering_passes): Do early local passes when called late.
4700 * tree-profile.c (do_tree_profiling): Don't profile functions added
4702 (do_early_tree_profiling, pass_early_tree_profile): Kill.
4703 * tree-cfg.c (update_modified_stmts): Do not update when operands are
4705 * passes.c (init_optimizations_passes): Reorder so we go into SSA
4706 during early_local_passes.
4707 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
4710 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4712 * Makefile.in: Update copyright year.
4714 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
4716 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
4717 in $(lang_checks) and check-consistency targets.
4719 2007-01-02 Jan Hubicka <jh@suse.cz>
4721 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
4722 Do not add referenced vars.
4723 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
4725 * passes.c (init_optimization_passes): Put mudflap_2 after
4726 free_datastructures.
4728 2007-01-02 Jan Hubicka <jh@suse.cz>
4730 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
4731 Set NOTHROW flag on call statements proved to be nothrow.
4732 Update statement of local calls so new pure/const functions are
4733 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
4735 (tree_rest_of_compilation): Register hooks and initialize bitmap
4736 early. Do not set after_inlining flag.
4738 2007-01-02 Steve Ellcey <sje@cup.hp.com>
4740 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
4741 HOST_BITS_PER_LONGLONG
4743 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4746 * c-typeck.c (store_init_value): Don't emit pedantic overflow
4747 warning for non-static initializers.
4749 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
4751 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
4752 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
4753 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
4754 Always use set_unique_reg_note to add REG_EQUAL notes.
4756 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4759 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4761 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4762 * cfgbuild.c (find_basic_blocks): Likewise.
4763 * cfgrtl.c (rtl_create_basic_block): Likewise.
4764 * function.c (temp_slots_at_level): Likewise.
4765 * reg-stack.c (stack_regs_mentioned): Likewise.
4766 * regclass.c (allocate_reg_info): Likewise.
4767 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4768 set_bb_for_stmt, move_block_to_fn): Likewise.
4769 * tree-complex.c (tree_lower_complex): Likewise.
4770 * vec.h (VEC_safe_grow_cleared): New.
4772 2007-01-02 Ian Lance Taylor <iant@google.com>
4774 * c-common.c (c_common_truthvalue_conversion): When warning about
4775 using an assignment as a truth value, set TREE_NO_WARNING.
4777 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4780 * c.opt (Wold-style-declaration): New.
4781 * doc/invoke.texi (C-only Warning Options): New.
4782 (Wold-style-declaration): Document it.
4783 (Wextra): Enabled by -Wextra.
4784 * c-opts.c (c_common_post_options): Enabled by -Wextra.
4785 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
4786 -Wold-style-declaration.
4788 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
4790 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
4791 * cfgbuild.c (find_basic_blocks): Likewise.
4792 * cfgrtl.c (rtl_create_basic_block): Likewise.
4793 * function.c (temp_slots_at_level): Likewise.
4794 * reg-stack.c (stack_regs_mentioned): Likewise.
4795 * regclass.c (allocate_reg_info): Likewise.
4796 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
4797 set_bb_for_stmt, move_block_to_fn): Likewise.
4798 * tree-complex.c (tree_lower_complex): Likewise.
4799 * vec.h (VEC_safe_grow_cleared): New.
4801 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
4803 * c-common.c (c_common_nodes_and_builtins): Since variants of
4804 void_type_node get built before it is given a name, we need to
4805 give those variants the name, too.
4806 (complete_array_type): We need to work with the canonical main
4807 type of the array, from which we will build the qualified version.
4808 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
4809 * print-tree.c (print_node): Display canonical type information
4811 * stor-layout.c (layout_type): When we don't know the
4812 alignment of a type for which we're building an array, we end up
4813 guessing wrong, so make the type require structural equality.
4814 * tree.c (make_node_stat): When we build a new type, it is its
4816 (build_type_attribute_qual_variant): When building an attribute
4817 variant, its canonical type is the non-attribute variant. However,
4818 if the attributes are target-dependent and they differ, we need to
4819 use structural equality checks for this type.
4820 (build_qualified_type): A qualified type is not equivalent to its
4821 unqualified variant; set the canonical type appropriately.
4822 (build_distinct_type_copy): When building a distinct type from
4823 another type, the new type is its own canonical type.
4824 (build_variant_type_copy): When building a new type variant, we
4825 assume that it is equivalent to the original type.
4826 (build_pointer_type_for_mode): When building a pointer type, also
4827 build a canonical type pointer.
4828 (build_reference_type_for_mode): When building a reference type,
4829 also build a canonical type reference.
4830 (build_index_type): When we can't hash an index type (e.g.,
4831 because its maximum value is negative), the index type requires
4832 structural equality tests.
4833 (build_array_type): Build the canonical form of an array type.
4834 (build_function_type): Function types require structural equality,
4835 because they contain default arguments, attributes, etc.
4836 (build_method_type_directly): Ditto for method types.
4837 (build_offset_type): Build the canonical offset type.
4838 (build_complex_type): Build the canonical vector type.
4839 (make_vector_type): Build the canonical vector type.
4840 * tree.h (TYPE_CANONICAL): New.
4841 (TYPE_STRUCTURAL_EQUALITY_P): New.
4842 (SET_TYPE_STRUCTURAL_EQUALITY): New.
4843 (struct tree_type): Added "canonical" field.
4844 * params.h (VERIFY_CANONICAL_TYPES): New.
4845 * doc/c-tree.texi (TYPE_CANONICAL): Document.
4846 (TYPE_STRUCTURAL_EQUALITY_P): Document.
4847 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
4848 * doc/invoke.texi (verify-canonical-types): Document --param
4849 parameter for verifying canonical types.
4851 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4853 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
4854 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
4855 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
4856 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
4858 * config/rs6000/e500.h: New.
4859 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
4860 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
4862 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
4863 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
4864 TARGET_E500_DOUBLE): Remove.
4865 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
4866 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
4867 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
4868 * config/rs6000/rs6000.c (rs6000_override_options): Use
4871 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4873 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
4874 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
4875 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
4876 rs6000_emit_cmove): Don't check TARGET_E500.
4877 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
4878 sordered): Don't check TARGET_E500.
4880 2007-01-01 Eric Christopher <echristo@apple.com>
4882 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
4883 HARD_REGISTER_NUM_P.
4885 2007-01-01 Roger Sayle <roger@eyesopen.com>
4887 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
4888 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
4889 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
4892 2007-01-01 Mike Stump <mrs@apple.com>
4894 * configure.ac: Remove support for building with Apple's gcc-3.1.
4896 2007-01-02 Joseph Myers <joseph@codesourcery.com>
4899 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
4900 REG is a hard register. Do not modify REG before calling
4902 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
4903 multiple of the size of YMODE for certain lowpart cases.
4905 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4908 * gimplify (voidify_wrapper_expr): Update for
4911 2007-01-01 Andreas Schwab <schwab@suse.de>
4914 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
4915 save of BR0 in extra_spill_size instead of spill_size.
4916 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
4918 (ia64_expand_epilogue): Restore BR0 from its new location.
4920 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4922 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
4923 Use a temporary variable if the left hand side is not a gimple
4926 2007-01-01 Andrew Pinski <pinskia@gmail.com>
4928 * gimplify.c (gimplify_return_expr): Make the temporary variable
4929 for the return expression, a gimple register variable.
4931 2007-01-01 Jan Hubicka <jh@suse.cz>
4933 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
4936 2007-01-01 Mike Stump <mrs@apple.com>
4938 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
4940 * config/darwin.c (machopic_select_rtx_section): Use
4942 (darwin_mergeable_constant_section): Likewise.
4943 * configure: Regenerate.
4944 * config.in: Regenerate.
4946 2007-01-01 Jan Hubicka <jh@suse.cz>
4947 Andrew Pinski <pinskia@gmail.com>
4949 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
4950 before starting IPA passes.