1 2007-01-28 Jan Hubicka <jh@suse.cz>
3 * builtins.c (expand_builtin_memset): Fix typo in my last patch.
4 * value-prof.c (stringop_block_profile): Likewise.
6 2007-01-28 Jan Hubicka <jh@suse.cz>
8 * expr.c (emit_block_move_via_movmem, emit_block_move_via_libcall): Add
9 variant handling histograms; add wrapper.
10 (clear_storage_via_libcall): Export.
11 (emit_block_move_hints): Break out from ...; add histograms.
12 (emit_block_move): ... this one.
13 (clear_storage_hints): Break out from ...; add histograms.
14 (clear_storage): ... this one.
15 (set_storage_via_memset): Handle histogram.
16 * expr.h (emit_block_move_via_libcall, emit_block_move_hints): Declare.
17 (clear_storage_hints, clear_storage_via_libcall): Declare.
18 (set_storage_via_setmem): Update prototype.
19 * doc/md.texi (movmem, setmem): Document new arguments.
21 * value-prof.c (dump_histogram_value, tree_find_values_to_profile): Add
23 (stringop_block_profile): New global function.
24 (tree_stringops_values_to_profile): Profile block size and alignment.
25 * value-prof.h (enum hist_type): add HIST_TYPE_AVERAGE and
27 (struct profile_hooks): Add gen_average_profiler and gen_ior_profiler.
28 (stringop_block_profile): Declare.
29 * builtins.c: Include value-prof.h.
30 (expand_builtin_memcpy, expand_builtin_memset): Pass block profile.
31 * gcov-ui.h (GCOV_COUNTER_NAMES): Add new counter.
32 (GCOV_COUNTER_AVERAGE, GCOV_COUNTER_IOR): New constants.
33 (GCOV_COUNTERS, GCOV_LAST_VALUE_COUNTER): Update.
34 * profile.c (instrument_values): Add new counters.
35 * cfgexpand.c (expand_gimple_basic_block): Propagate histograms to
37 * tree-profile.c (tree_average_profiler_fn, tree_ior_profiler_fn): New.
38 (tree_init_edge_profiler): Build new profilers.
39 (tree_gen_average_profiler, tree_gen_ior_profiler): New.
40 (pass_tree_profile): Add dump.
41 (tree_profile_hooks): Update.
42 * Makefile.in (LIBGCOV): Add new constants.
43 * libgcov.c (__gcov_merge_ior, __gcov_average_profiler,
44 __gcov_ior_profiler): New.
45 * i386.md (movmem/setmem expanders): Add new optional arguments.
47 2007-01-28 David Edelsohn <edelsohn@gnu.org>
49 * doc/md.texi (Standard Pattern Names): Document blockage pattern.
51 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
53 * tree-ssa-loop-unswitch.c: Include tree-inline.h.
54 (tree_unswitch_single_loop): Pass eni_size_weights to
56 * tree-ssa-loop-manip.c: Include tree-inline.h.
57 (can_unroll_loop_p): Pass eni_size_weights to
59 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
60 Pass eni_size_weights to estimate_num_insns.
61 * tree.h (init_inline_once): Export.
62 * toplev.c (backend_init): Call init_inline_once.
63 * cgraphunit.c (cgraph_process_new_functions,
64 cgraph_analyze_function): Pass eni_inlining_weights to
66 * ipa-inline.c (compute_inline_parameters): Ditto.
67 * tree-ssa-loop-ivcanon.c (tree_num_loop_insns): Pass weights
68 to estimate_num_insns.
69 (try_unroll_loop_completely): Pass eni_size_weights to
71 * tree-eh.c (decide_copy_try_finally): Pass eni_size_weights
72 ot estimate_num_insns.
73 * tree-ssa-loop-prefetch.c: Include tree-inline.h.
74 (loop_prefetch_arrays): Pass eni_time_weights to tree_num_loop_insns.
75 * tree-inline.c (eni_inlining_weights, eni_size_weights,
76 eni_time_weights): New variables.
77 (init_inline_once): Initialize them.
78 (struct eni_data): Mew.
79 (estimate_num_insns_1, estimate_num_insns): Use weights.
80 * tree-inline.h (struct eni_weights_d): New.
81 (eni_inlining_weights, eni_size_weights, eni_time_weights): Declare.
82 (estimate_num_insns): Declaration changed.
83 * cfgloop.h (tree_num_loop_insns): Declaration changed.
84 * Makefile.in (tree-ssa-loop-unswitch.o, tree-ssa-loop-prefetch.o,
85 tree-ssa-loop-manip.o): Add TREE_INLINE_H dependency.
87 2007-01-28 Zdenek Dvorak <dvorakz@suse.cz>
89 * tree-data-ref.c (conflict_fn): Assert that the number of affine
90 relations in the conflict function is valid.
92 2007-01-27 Ian Lance Taylor <iant@google.com>
94 * common.opt: Add fstrict-overflow.
95 * opts.c (decode_options): Set flag_strict_overflow if -O2.
96 * flags.h (TYPE_OVERFLOW_WRAPS): Define.
97 (TYPE_OVERFLOW_UNDEFINED): Define.
98 (TYPE_OVERFLOW_TRAPS): Define. This replaces TYPE_TRAP_SIGNED.
100 * tree.h (TYPE_TRAP_SIGNED): Don't define.
101 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_UNDEFINED.
102 (fold_negate_expr): Likewise.
103 (make_range): Likewise.
104 (extract_muldiv_1): Likewise.
105 (maybe_canonicalize_comparison): Likewise.
106 (fold_comparison): Likewise.
107 (fold_binary): Likewise.
108 (tree_expr_nonnegative_p): Likewise.
109 (tree_expr_nonzero_p): Likewise.
110 * tree-vrp.c (compare_values): Likewise.
111 (extract_range_from_binary_expr): Likewise.
112 (extract_range_from_unary_expr): Likewise.
113 * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness):
115 (nowrap_type_p): Likewise.
116 * tree-scalar-evolution.c (simple_iv): Likewise.
117 * fold-const.c (negate_expr_p): Use TYPE_OVERFLOW_WRAPS.
118 (build_range_check): Likewise.
119 (extract_muldiv_1): Likewise.
120 (fold_comparison): Likewise.
121 * tree-vrp.c (vrp_int_const_binop): Likewise.
122 (extract_range_from_unary_expr): Likewise.
123 * convert.c (convert_to_integer): Likewise.
124 * fold-const.c (fold_negate_expr): Use TYPE_OVERFLOW_TRAPS.
125 (fold_comparison): Likewise.
126 (fold_binary): Likewise.
127 * optabs.c (optab_for_tree_code): Likewise.
128 * tree-vectorizer.c (vect_is_simple_reduction): Likewise.
129 * simplify-rtx.c (simplify_const_relational_operation): Check
130 flag_strict_overflow and flag_trapv.
131 (simplify_const_relational_operation): Likewise.
132 * doc/invoke.texi (Option Summary): Mention -fstrict-overflow.
133 (Optimize Options): Add -fstrict-overflow to -O2 list. Document
136 2007-01-27 Roger Sayle <roger@eyesopen.com>
138 * tree.c (tree_fold_gcd): Delete.
139 * tree.h (tree_fold_gcd): Remove prototype.
140 * tree-data-ref.c (tree_fold_divides_p): Don't use tree_fold_gcd to
141 test whether one constant integer is a multiple of another. Instead
142 call int_const_binop with TRUNC_MOD_EXPR and test for a zero result.
143 * fold-const.c (multiple_of_p): We've determined both TOP and
144 BOTTOM are integer constants so we can call int_const_binop directly
145 instead of the more generic const_binop.
147 2007-01-27 Roger Sayle <roger@eyesopen.com>
149 * fold-const.c (size_binop): In the fast-paths for X+0, 0+X, X-0 and
150 1*X check that the constant hasn't overflowed, to preserve the
152 (round_up): Provide an efficient implementation when rouding-up an
153 INTEGER_CST to a power-of-two.
155 2007-01-28 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
157 * doc/sourcebuild.texi: Add comma for clarity.
158 * doc/extend.texi: Fix some typos.
159 * doc/passes.texi: Likewise.
160 * doc/cppinternals.texi: Likewise.
161 * doc/c-tree.texi: Likewise.
162 * doc/tree-ssa.texi: Likewise.
163 * doc/install.texi: Likewise.
165 2007-01-27 Jan Hubicka <jh@suse.cz>
167 * tree-sra.c (sra_walk_function): Don't rely on aliases being build.
168 (pass_sra): Do not require alias information.
169 * passes.c (init_optimization_passes): Add SRA
171 2007-01-27 Steven Bosscher <steven@gcc.gnu.org>
173 * tracer.c (rest_of_handle_tracer): We already cleaned
174 up the CFG in tracer() so don't do it here again.
175 * cfgcleanup.c (rest_of_handle_jump2): Don't repeat
176 cleanup_cfg here, either. And don't call renumber_insns.
178 * cfgrtl.c (rtl_verify_flow_info_1): Don't verify that BB_END
179 and BB_HEAD are in the insn stream here. Instead make sure
180 that BB_INSN is valid on all insns. Also, do check here that
181 there are no pending branch predictions...
182 (rtl_verify_flow_info): ...instead of doing it here. Checks
183 for BB_END and BB_HEAD moved from rtl_verify_flow_info_1 to
186 2007-01-26 Roger Sayle <roger@eyesopen.com>
188 * config/i386/i386.c (ix86_swap_binary_operands_p): New helper
189 function to simplify/factorize operand order canonicalization.
190 (ix86_fixup_binary_operands): Reorganize using the above function.
191 (ix86_binary_operator_ok): Likewise.
193 2007-01-27 Jakub Jelinek <jakub@redhat.com>
195 * genattrtab.c (struct attr_value_list, insn_code_values): Move to
196 file scope from optimize_attrs.
197 (simplify_test_exp): If insn_code_values is not NULL, use it to speed
199 (optimize_attrs): Clear insn_code_values after freeing it.
201 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
203 * tree-ssa-address.c (create_mem_ref): Remove ", bsi" from
204 a parts.base assignment.
206 2007-01-26 Zdenek Dvorak <dvorakz@suse.cz>
208 * tree-data-ref.c (dump_subscript): Use dump_conflict_function.
209 (compute_subscript_distance, initialize_data_dependence_relation,
210 finalize_ddr_dependent, analyze_ziv_subscript,
211 analyze_siv_subscript_cst_affine,
212 compute_overlap_steps_for_affine_univar,
213 compute_overlap_steps_for_affine_1_2, analyze_subscript_affine_affine,
214 analyze_siv_subscript, analyze_miv_subscript,
215 analyze_overlapping_iterations, subscript_dependence_tester_1,
216 compute_self_dependence, free_dependence_relation): Work
217 with affine_fn instead of chrecs.
218 (dump_affine_function, dump_conflict_function, affine_function_equal_p,
219 common_affine_function, affine_function_base,
220 affine_function_constant_p, affine_fn_op, affine_fn_plus,
221 affine_fn_minus, affine_fn_free, conflict_fn_not_known,
222 conflict_fn_no_dependence, free_conflict_function, free_subscripts,
223 conflict_fn, affine_fn_cst, affine_fn_univar): New functions.
224 (all_chrecs_equal_p): Removed.
225 * tree-data-ref.h (affine_fn, conflict_function): New types.
226 (struct subscript): Change type of conflicting_iterations_in_a
227 and conflicting_iterations_in_b.
229 2007-01-26 Steve Ellcey <sje@cup.hp.com>
232 * config/pa/pa.h (TARGET_HPUX_11): New.
233 * config/pa/pa-hpux11.h (TARGET_HPUX_11): New.
234 * config/pa/pa.c (pa_init_builtins): Use TARGET_HPUX_11.
236 2007-01-26 Daniel Berlin <dberlin@dberlin.org>
237 Richard Guenther <rguenther@suse.de>
239 * tree-ssa-structalias.c (solve_graph): Handle case
240 we merged the variable to another.
242 2007-01-25 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
244 * builtins.c (fold_builtin_1): Treat ccos and ccosh as 'even'
247 * fold-const.c (negate_mathfn_p): Treat casin, casinh, catan,
248 catanh, cproj, csin, csinh, ctan and ctanh as 'odd' functions.
250 2007-01-25 DJ Delorie <dj@redhat.com>
252 * config/m32c/m32c.c (m32c_cannot_change_mode_class): We don't
253 allow changes to modes which don't fit in those registers.
255 * reload1.c (choose_reload_regs): Check for invalid subregs before
256 computing their locations, not after.
258 2007-01-25 Geoffrey Keating <geoffk@apple.com>
261 * config/rs6000/rs6000.c (first_altivec_reg_to_save): On Darwin,
262 save Altivec registers in an eh_return function.
263 (compute_vrsave_mask): Likewise.
264 (rs6000_stack_info): Correct AIX/Darwin stack alignment computation
265 for saving Altivec registers.
266 (rs6000_emit_prologue): Don't allocate stack twice in
267 eh_return function. Correct expected value of altivec_save_offset
268 when using save_world. Describe save of R0 to stack when using
269 save_world. Describe stack pointer adjustment when using
270 save_world. Remove duplicated eh_return parameter register saving.
271 Update sp_offset variable after save_world.
272 * config/rs6000/t-darwin (LIB2FUNCS_STATIC_EXTRA): Remove
274 (LIB2FUNCS_EXTRA): Add darwin-world.asm.
275 * config/rs6000/darwin.h (SUBTARGET_OVERRIDE_OPTIONS): -m64
278 2007-01-25 Steve Ellcey <sje@cup.hp.com>
280 * config.gcc (ia64*-*-hpux*): Make posix threads the default.
282 2007-01-25 Steve Ellcey <sje@cup.hp.com>
285 * config/pa/pa.c (pa_init_builtins): Set asm names for finite routines.
286 * config/ia64/ia64.c (ia64_init_builtins): Ditto.
288 2007-01-25 Richard Guenther <rguenther@suse.de>
290 * doc/invoke.texi (-Wcoverage-mismatch): Document.
291 * common.opt (-Wcoverage-mismatch): New warning option.
292 * coverage.c (get_coverage_counts): Ignore coverage mismatch
293 if -Wcoverage-mismatch is given.
295 2007-01-25 Razya Ladelsky <razya@il.ibm.com>
297 * ipa-cp.c (ipcp_insert_stage, ipcp_driver): Support for SSA.
298 (ipcp_driver): Change to static definition.
299 Add dumping of the ifunctions.
300 (constant_val_insert): Remove unused parameter. Support for SSA.
301 (ipcp_propagate_const): Support for SSA.
302 (ipcp_profile_bb_print): Print only analyzed nodes.
303 (ipcp_replace_map_create): Remove support for Fortran constant
305 * ipa-prop.c (ipa_method_modify_stmt,
306 ipa_callsite_compute_param): Support for SSA.
307 * ipa-prop.h (ipcp_driver): Remove declaration.
308 (IS_VALID_TREE_MAP_INDEX): Add define.
310 2007-01-24 Geoffrey Keating <geoffk@apple.com>
312 * unwind-dw2.c (execute_stack_op): Handle DW_OP_swap.
314 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
317 * builtins.c (fold_builtin_cabs): Use MPFR to evaluate a
318 constant argument to cabs and do it without checking for
319 -funsafe-math-optimizations.
321 2007-01-24 Douglas Gregor <dgregor@osl.iu.edu>
323 * c-common.h (RID_FIRST_CXX0X): New.
324 (RID_LAST_CXX0X): New.
325 * c-opts.c (c_common_handle_option): -Wc++0x-compat is triggered
327 * c.opt (Wc++0x-compat): New.
328 * doc/invoke.texi (-Wc++0x-compat): Document.
330 2007-01-24 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
332 * builtins.c (fold_builtin_carg): New.
333 (fold_builtin_1): Use it.
335 2007-01-24 Jan Hubicka <jh@suse.cz>
337 * ipa-inline.c (cgraph_decide_inlining): Initialize initial_insns.
339 * ipa-inline.c (initial_insns, max_insns): Delete.
340 (compute_max_insns): New function.
341 (cgraph_decide_inlining_of_small_function): Use it; take minimal amount
342 of insns as base for code growth.
343 (cgraph_decide_inlining): Make initial_insns local; do not compute
345 * params.def (PARAM_INLINE_UNIT_GROWTH): Set to 60.
346 * doc/invoke.texi (inline-unit-growth): Update docs.
348 2007-01-24 Jakub Jelinek <jakub@redhat.com>
350 * config/i386/i386.h (x86_cmpxchg16b): Remove const.
351 (TARGET_CMPXCHG16B): Define to x86_cmpxchg16b.
352 * config/i386/i386.c (x86_cmpxchg16b): Remove const.
353 (override_options): Add PTA_CX16 flag. Set x86_cmpxchg16b
354 for CPUs that have PTA_CX16 set.
357 * gimplify.c (omp_check_private): New function.
358 (gimplify_scan_omp_clauses): Use it for
359 firstprivate/lastprivate/reduction.
362 * gimplify.c (omp_add_variable): Don't call omp_notice_variable
363 on TYPE_SIZE_UNIT for GOVD_LOCAL VLAs.
366 * omp-low.c (lower_omp_for_lastprivate): Add dlist argument.
367 If lower_lastprivate_clauses emits some statements, append them
368 to dlist rather than body_p and to body_p append an initializer.
369 (lower_omp_for): Adjust caller.
371 2007-01-24 Steve Ellcey <sje@cup.hp.com>
373 * target.h (globalize_decl_name): New.
374 * target-def.h (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
375 * output.h (default_globalize_decl_name): New.
376 * varasm.c (asm_output_bss): Use globalize_decl_name instead of
378 (globalize_decl): Ditto.
379 (default_globalize_decl_name): New.
380 * config/ia64/ia64.c (ia64_globalize_decl_name): New.
381 (ia64_handle_version_id_attribute): New.
382 (TARGET_ASM_GLOBALIZE_DECL_NAME): New.
383 (ia64_asm_output_external): Use globalize_decl_name instead
385 * doc/extend.texi (version_id): New pragma.
386 * doc/tm.texi (ARGET_ASM_GLOBALIZE_DECL_NAME): New target hook.
388 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
390 * unwind-dw2-fde.c (get_cie_encoding): Replaced _Unwind_Word with
391 _uleb128_t and _Unwind_SWord with _sleb128_t.
392 * unwind-dw2.c (extract_cie_info, execute_stack_op, execute_cfa_program,
393 uw_frame_state_for, uw_update_context_1): Likewise.
394 * unwind-c.c (parse_lsda_header, PERSONALITY_FUNCTION): Likewise.
395 * unwind-pe.h (read_uleb128, read_sleb128,
396 read_encoded_value_with_base): Likewise.
397 * unwind-generic.h: Define _sleb128_t and _uleb128_t types.
399 2007-01-24 Richard Guenther <rguenther@suse.de>
401 * builtins.c (expand_builtin_cexpi): Get the fndecl
402 for cexp in the correct way.
404 2007-01-24 Jan Hubicka <jh@suse.cz>
406 * tree-ssa-dce.c (eliminate_unnecesary_stmts): Remove dead LHS of calls.
408 2007-01-24 Andreas Krebbel <krebbel1@de.ibm.com>
410 * c-cppbuiltin.c (builtin_define_type_sizeof): New function.
411 (c_cpp_builtins): New builtin macros: __SIZEOF_INT__, __SIZEOF_LONG__,
412 __SIZEOF_LONG_LONG__, __SIZEOF_SHORT__, __SIZEOF_POINTER__,
413 __SIZEOF_FLOAT__, __SIZEOF_DOUBLE__, __SIZEOF_LONG_DOUBLE__,
414 __SIZEOF_SIZE_T__, __SIZEOF_WCHAR_T__, __SIZEOF_WINT_T__ and
415 __SIZEOF_PTRDIFF_T__.
416 * doc/cpp.texi: Documentation for the new builtin macros added.
418 2007-01-24 Uros Bizjak <ubizjak@gmail.com>
420 * config/i386/i386.md (tanxf2, tan<mode>2, atan<mode>2, log<mode>2,
421 log10<mode>2, log2<mode>2, expxf2, exp10xf2, exp2xf2): Use op2
422 instead of operands[2] to avoid access past the end of array.
424 2007-01-24 Richard Sandiford <richard@codesourcery.com>
426 * reload1.c (emit_reload_insns): Pass the reload register
427 for a non-spill output reload through forget_old_reloads_1.
429 2007-01-23 Joseph Myers <joseph@codesourcery.com>
431 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Define
432 _SOFT_DOUBLE if doubles use software floating-point.
433 * config/rs6000/libgcc-ppc-glibc.ver: Export additional long
434 double functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
435 * config/rs6000/darwin-ldouble.c: Also compile functions for
436 hard-float without FPRs. Use fmsub function for all __NO_FPRS__
437 cases. Compile extra functions if _SOFT_DOUBLE, not _SOFT_FLOAT.
438 * config/rs6000/linuxspe.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Remove
439 commented-out long double override.
440 (CPP_LONGDOUBLE_DEFAULT_SPEC): Likewise.
441 * config/rs6000/eabispe.h: Likewise.
442 * config/rs6000/rs6000.c (rs6000_override_options): Don't override
443 long double for non-SPE.
444 (rs6000_handle_option): Likewise.
445 (invalid_e500_subreg): Disallow more subregs involding DImode,
446 DFmode, TImode or TFmode.
447 (rs6000_legitimate_offset_address_p): Check TFmode offsets for
449 (legitimate_lo_sum_address_p): Also check for TFmode for E500
451 (rs6000_legitimize_address): Also handle TFmode for E500 double.
452 (rs6000_legitimize_reload_address): Also handle TFmode for E500
454 (rs6000_legitimate_address): Also check for TFmode for E500
456 (rs6000_emit_move): Use DFmode subregs of TFmode for E500 double.
457 (spe_build_register_parallel): Handle TFmode and TCmode.
458 (rs6000_spe_function_arg): Handle TFmode and TCmode for E500
460 (function_arg): Handle TFmode and TCmode for E500 double.
461 (rs6000_init_libfuncs): Initialize extra libfuncs for soft double
463 (print_operand): Handle TFmode and TImode for %y.
464 (rs6000_generate_compare): Handle TFmode comparisons for E500
466 (spe_func_has_64bit_regs_p): Check for TFmode for E500 double.
467 (rs6000_function_value): Handle TFmode and TCmode for E500 double.
468 (rs6000_libcall_value): Handle TFmode and TCmode for E500 double.
469 * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Check for
470 TFmode for E500 double.
471 * config/rs6000/rs6000.md (FP): Allow TF for E500 double.
472 (floatsidf2): Enable for E500 double.
473 (movtf_softfloat): Use rs6000_nonimmediate_operand.
474 (extenddftf2): Change to extenddftf2_fprs.
475 (extenddftf2): Call gen_spe_extenddftf2 or gen_extenddftf2_fprs
476 depending on TARGET_E500_DOUBLE.
477 (extendsftf2): Enable for E500 double.
478 (trunctfdf2): Enable for E500 double.
479 (trunctfsf2): Change to trunctfsf2_fprs.
480 (trunctfsf2): Call gen_spe_trunctfsf2 or gen_trunctfsf2_fprs
481 depending on TARGET_E500_DOUBLE.
482 (floatsitf2): Enable for E500 double.
483 (fix_trunctfsi2): Change to fix_trunctfsi2_fprs.
484 (fix_trunctfsi2): Call gen_spe_fix_trunctfsi2 or
485 gen_fix_trunctfsi2_fprs depending on TARGET_E500_DOUBLE.
486 (negtf2): Change to negtf2_internal.
487 (negtf2): New expander.
488 (abstf2): Enable for E500 double. Call gen_spe_abstf2_tst,
489 gen_spe_abstf2_cmp or gen_abstf2_internal depending on
490 TARGET_E500_DOUBLE and flag_unsafe_math_optimizations.
491 (movdi_internal32): Use rs6000_nonimmediate_operand.
492 (unnamed splitter): Likewise.
493 * config/rs6000/spe.md (CMPTFEQ_GPR, TSTTFEQ_GPR, CMPTFGT_GPR,
494 TSTTFGT_GPR, CMPTFLT_GPR, TSTTFLT_GPR): New unspecs.
495 (SPE64TF, DITI): New mode macros.
496 (frob_df_di): Change to frob_<SPE64:mode>_<DITI:mode>; allow more
499 (frob_<mode>_di_2): New.
500 (frob_tf_di_8_2): New.
501 (frob_di_df): Change to frob_di_<mode>; allow more modes.
503 (frob_di_df_2): Change to frob_<DITI:mode>_<SPE64:mode>_2; allow
505 (frob_ti_<mode>_8_2): New.
507 (mov_si<mode>_e500_subreg0, mov_si<mode>_e500_subreg0_2,
508 mov_si<mode>_e500_subreg4, mov_si<mode>_e500_subreg4_2): Allow
510 (mov_sitf_e500_subreg8, mov_sitf_e500_subreg8_2,
511 mov_sitf_e500_subreg12, mov_sitf_e500_subreg12_2): New.
512 (spe_trunctfdf2_internal1, spe_trunctfsf2, spe_extenddftf2,
513 spe_fix_trunctfsi2, spe_fix_trunctfsi2_internal,
514 spe_negtf2_internal, spe_abstf2_cmp, spe_abstf2_tst): New.
515 (cmptfeq_gpr, tsttfeq_gpr, cmptfgt_gpr, tsttfgt_gpr, cmptflt_gpr,
518 2007-01-23 Ian Lance Taylor <iant@google.com>
520 * Makefile.in (OBJS-common): Reformat, alphabetize, but put
522 (OBJS-archive): Reformat, alphabetize.
523 (OBJS): Change out_object_file to OBJS-md.
525 2007-01-23 Uros Bizjak <ubizjak@gmail.com>
527 * config/i386/i386.md: Use REG_P, MEM_P, CONST_INT_P, LABEL_P,
528 JUMP_P and CALL_P predicates where applicable.
529 * config/i386/i386.c: Ditto.
530 * config/i386/i386.md: Ditto.
531 * config/i386/mmx.md: Ditto.
532 * config/i386/predicates.md: Ditto.
534 2007-01-22 Andreas Schwab <schwab@suse.de>
536 * config/m68k/m68k.h: Fix comment.
538 2007-01-22 Jan Hubicka <jh@suse.cz>
540 * passes.c (init_optimization_passes): Do not rerun
541 pass_early_warn_uninitialized.
543 2007-01-22 Richard Guenther <rguenther@suse.de>
545 PR tree-optimization/30038
546 * tree-ssa-math-opts.c (maybe_record_sincos): New static helper
548 (execute_cse_sincos_1): Likewise.
549 (execute_cse_sincos): Likewise.
550 (gate_cse_sincos): Likewise.
551 (pass_cse_sincos): New pass CSEing sin() and cos() calls using
552 the cexpi() canonicalization of sincos().
553 * tree-pass.h (pass_cse_sincos): Declare.
554 * passes.c (init_optimization_passes): New pass pas_cse_sincos.
556 2007-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
558 PR rtl-optimization/29329
559 * combine.c (replaced_rhs_insn): Rename to i2mod.
560 (replaced_rhs_value): Rename to i2mod_new_rhs.
561 (i2mod_old_rhs): New global variable.
562 (combine_instructions): Adjust for above change. Save a copy of
563 the old RHS into i2mod_old_rhs when the contents of a REG_EQUAL
564 note are substituted in the second instruction.
565 (distribute_notes) <REG_DEAD>: Adjust for above change. Do not
566 ditch the note if it pertains to the second eliminated register
567 and this register is mentioned in i2mod_old_rhs.
570 2006-09-12 Eric Botcazou <ebotcazou@libertysurf.fr>
572 * combine.c (distribute_notes) <REG_DEAD>: Do not consider SETs past
573 the insn to which the note was originally attached.
575 2007-01-21 Jan Hubicka <jh@suse.cz>
577 * ipa-inline.c (inlining_mode): Comment, move up.
578 (cgraph_decide_inlining_incrementally): Do not perform inlining
579 itself; fix handling of flattening of self recursive functions.
580 (cgraph_find_cycles): Remove.
581 (cgraph_flatten_node): Remove.
582 (cgraph_decide_inlining): Use incremental inliner to handle flattening.
583 (try_inline): New function.
584 (cgraph_early_inlining): Update call of
585 cgraph_decide_inlining_incrementally. Apply inlining here.
586 (apply_inline): Update call of cgraph_decide_inlining_incrementally.
588 2007-01-21 Dirk Mueller <dmueller@suse.de>
591 * tree-vrp.c (check_array_bounds): do not warn
592 about ADDR_EXPR's of ARRAY_REF's which are immediately
593 used in binary expressions.
595 2007-01-21 Ira Rosen <irar@il.ibm.com>
597 * tree-vectorizer.h (struct _stmt_vec_info): Add new field
598 read_write_dep and macros for its access.
599 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
600 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Remove
601 argument, call vect_check_interleaving for every independent pair of
602 data-refs. Mark loads that access the same memory location as a store
604 (vect_check_dependences): Remove.
605 (vect_analyze_data_ref_dependences): Remove vect_check_dependences
606 call, fix the call to vect_analyze_data_ref_dependence.
607 (vect_analyze_data_ref_access): For statements that access the same
608 data-ref, check that they are not stores; for loads, check that there
609 is no store that access the same location.
611 2007-01-20 Roger Sayle <roger@eyesopen.com>
612 Joseph Myers <joseph@codesourcery.com>
614 * doc/invoke.texi (-fdump-rtl-combine): Fix under/overfull hbox.
615 (-fdump-rtl-stack): Likewise.
616 (-fno-signed-zeros): Use @minus{} for a minus sign. Correct typo.
617 (-mcheck-zero-division, -mcpu): Fix under/overfull hbox.
618 (-mpt-fixed): Use @minus{} for minus sign.
619 (Using Precompiled Headers): Fix under/overfull hbox.
621 2007-01-20 Jan Hubicka <jh@suse.cz>
623 * tree-flow.h (struct stmt_ann_d): Move references_memory to proper
624 place within annotation.
626 2007-01-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
628 * pa.c (output_move_double): Change array size of xoperands to 4.
630 2007-01-20 Jan Hubicka <jh@suse.cz>
632 * tree-tailcall.c (adjust_return_value): Do not use RESULT_DECL
635 2007-01-19 Ian Lance Taylor <iant@google.com>
637 * expmed.c (expand_divmod) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]: Cast
638 constant to unsigned HOST_WIDE_INT before negating.
640 2007-01-19 Ian Lance Taylor <iant@google.com>
642 * tree-ssa-operands.h (struct vuse_vec_d): Change num_vuse field
644 (VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Use unsigned comparison.
645 (VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
646 (SET_VUSE_VECT_ELEMENT) [ENABLE_CHECKING]: Likewise.
647 (SET_VUSE_ELEMENT_VAR) [ENABLE_CHECKING]: Likewise.
648 (SET_VUSE_ELEMENT_PTR) [ENABLE_CHECKING]: Likewise.
649 (realloc_vdef, realloc_vuse): Change second parameter to
651 (ssa_operand_iterator_d): Change vuse_index and mayuse_index
653 * tree-ssa-operands.c (realloc_vop): Change num_elem parameter to
654 unsigned. Change x and lim locals to unsigned.
655 (realloc_vdef, realloc_vuse): Change num_elem parameter to
657 (finalize_ssa_vuse_ops): Change old_i local to unsigned.
658 (copy_virtual_operands): Change i and n locals to unsigned.
660 2007-01-19 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
663 * builtins.c (fold_builtin_1): Handle builtin fdim.
665 2007-01-20 Jan Hubicka <jh@suse.cz>
667 * tree-ssa.c (init_tree_ssa): Do not call init_alias_heapvars.
668 * tree-ssa-structalias.c (compute_points_to_sets): Do call
670 (init_alias_heapvars): Initialize only when not already initialized.
671 (delete_alias_heapvars): Set heapvar_for_stmt to NULL.
673 2007-01-19 Roger Sayle <roger@eyesopen.com>
675 * common.opt (fsigned-zeros): New command line option.
676 * flags.h (HONOR_SIGNED_ZEROS): Control via flag_signed_zeros instead
677 of flag_unsafe_math_optimizations.
678 * opts.c (set_fast_math_flags): The -ffast-math command line option
679 implies -fno-signed-zeros.
680 (fast_math_flags_set_p): Likewise.
682 * doc/invoke.texi: Document new -fno-signed-zeros option, and update
683 the documentation of -ffast-math appropriately. Wrap long lines.
685 2007-01-19 Steve Ellcey <sje@cup.hp.com>
687 * system.h (ASM_MAKE_LABEL_LINKONCE): Poison.
688 * varasm.c (globalize_decl): Remove ASM_MAKE_LABEL_LINKONCE ifdef.
690 2007-01-19 Tomas Bily <tbily@suse.cz>
692 * cgraphunit.c (cgraph_finalize_function): Updating of pid
694 (tree_init_ic_make_global_vars): New function
695 (tree_init_edge_profiler): call of tree_init_ic_make_global_vars
696 (tree_gen_ic_profiler): New function
697 (tree_gen_ic_func_profiler): New function
698 (tree_profiling): Added calling of tree_gen_ic_func_profiler
699 (tree_profile_hooks): Added hook for indirec/virtual calls
700 * value-prof.c (tree_find_values_to_profile): New case for
702 (tree_values_to_profile): Call for determining indirect/virtual
704 (tree_indirect_call_to_profile): New function
705 (tree_ic_transform): New function
706 (tree_ic): New function
707 (find_func_by_pid): New function
708 (init_pid_map): New function
709 (tree_value_profile_transformations): Added check for
710 indirect/virtual call transformation
711 * value-prof.h (enum hist_type): New counter type for
712 indirect/virtual calls
713 (profile_hooks): Added new hook for profiling indirect/virtual
715 * profile.c (instrument_values): New case for indirect/virtual
717 * gcov-io.h (GCOV_LAST_VALUE_COUNTER): Changed to 6
718 (GCOV_COUNTER_V_INDIR): New counter type
719 (GCOV_COUNTER_NAMES): New name of counter "indirect" added
720 (GCOV_MERGE_FUNCTIONS): New merge function for indirect/virtual
722 * cgraph.c: Definition of cgraph_max_pid
723 (cgraph_create_node): Default init of pid attribute
724 * cgraph.h: Declaration of cgraph_max_pid
725 (struct cgraph_node): Added pid attribute
726 * libgcov.c (__gcov_indirect_call_profiler): New function
727 (__gcov_one_value_profiler_body): New function
728 (__gcov_one_value_profiler): Body was moved to
729 __gcov_one_value_profiler_body and calls it
731 2007-01-19 Basile Starynkevitch <basile@starynkevitch.net>
733 * doc/gty.texi (Options): Document the mark_hook option to GTY.
734 * gengtype.c (write_types_data, write_func_for_structure,
735 write_types, ggc_wtd, pch_wtd): Add skip_hooks to
736 write_types_data, ggc_wtd, pch_wtd for processing mark_hook.
737 (walk_type, write_func_for_structure): Generate the mark_hook if
740 2007-01-19 Jan Hubicka <jh@suse.cz>
742 * ipa-inline.c (cgraph_decide_inlining_incrementally): Instead of
743 'early' argument take inlining mode argument specifying whether to
744 inline for size/speeed or all functions; add support for flattening;
746 (cgraph_early_inlining): Update call of decide_inlining_incrementally.
748 2007-01-19 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
751 * toplev.c (warn_deprecated_use): Use %qD instead of %qs to print
752 the name of the declared identifier.
754 2007-01-19 Dirk Mueller <dmueller@suse.de>
756 * config/i386.h (CONDITIONAL_REGISTER_USAGE): Store
757 result of PIC_OFFSET_TABLE_REGNUM in temporary variable to avoid
758 duplicate evaluation.
760 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
762 * config/i386/i386.md (acos<mode>2): Rename from acossf2 and acosdf2.
763 Macroize expander using X87MODEF12 mode macro. Extend operand 1
764 to XFMode, use acosxf2 and truncate result to requested mode.
765 Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
766 (asin<mode>2): Similarly, with asin expanders.
767 (*fscalexf4_i387): Rename from *fscalexf4.
768 (expNcorexf3): New expander.
769 (expxf2, exp10xf2, exp2xf2): Use expNcorexf3 expander.
770 (exp<mode>2): Rename from expsf2 and expdf2. Macroize expander using
771 X87MODEF12 mode macro. Extend operand 1 to XFMode, use expxf2 and
772 truncate result to requested mode. Use SSE_FLOAT_MODE_P to disable
773 patterns for SSE math.
774 (exp10<mode>2): Similarly, with exp10 expanders.
775 (exp2<mode>2): Similarly, with exp2 expanders.
776 (expm1<mode>2): Similarly, with expm1 expanders.
777 (ldexp<mode>3): Similarly, with ldexp expanders.
778 (log<mode>2, log10<mode>2, log2<mode>2, log1p<mode>2, logb<mode>2):
779 Use gen_truncxf<mode>2_i387_noop to truncate result.
781 2007-01-19 Richard Sandiford <richard@codesourcery.com>
783 * config/m68k/m68k.h (PREFERRED_STACK_BOUNDARY): Define to 32
784 for ColdFire targets.
786 2007-01-19 Nathan Sidwell <nathan@codesourcery.com>
787 Richard Sandiford <richard@codesourcery.com>
789 * config/m68k/m68k.h (M68K_STATIC_CHAIN_REG_NAME): New macro.
790 (INITIALIZE_TRAMPOLINE): Use STATIC_CHAIN_REGNUM.
791 (__transfer_from_trampoline): Use M68K_STATIC_CHAIN_REG_NAME.
792 * config/m68k/m68kelf.h (STATIC_CHAIN_REGNUM): Override.
793 (M68K_STATIC_CHAIN_REG_NAME): Likewise.
794 * config/m68k/netbsd-elf.h (M68K_STATIC_CHAIN_REG_NAME): Likewise.
796 2007-01-19 Richard Sandiford <richard@codesourcery.com>
798 * config/m68k/m68k.md (adddi_dilshr32): Rename to...
799 (*adddi_dilshr32): ...this. Fix formatting. Remove commented-out
800 non-canonical pattern. Restrict to !TARGET_COLDFIRE.
801 (*adddi_dilshr32_cf): New pattern.
802 (adddi3, subdi3): Remove first alternatives.
804 2007-01-19 Richard Sandiford <richard@codesourcery.com>
806 * config/m68k/m68k.c (notice_update_cc): If an SFmode move is
807 implemented using move.l, do not use its cc result for floating-point
810 2007-01-19 Richard Sandiford <richard@codesourcery.com>
812 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Stop the 'T' constraint
813 from accepting 's' constraints if flag_pic.
815 2007-01-19 Richard Sandiford <richard@codesourcery.com>
817 * config/m68k/m68k.md (bordered, bunordered, buneq, bunge, bungt)
818 (bunle, bunlt, bltgt, bordered_rev, bunordered_rev, buneq_rev)
819 (bunge_rev, bungt_rev, bunle_rev, bunlt_rev, bltgt_rev): Change
820 condition from TARGET_68881 to TARGET_HARD_FLOAT.
822 2007-01-19 Sandra Loosemore <sandra@codesourcery.com>
824 * longlong.h (count_leading_zeros, COUNT_LEADING_ZEROS_0): Add
825 ColdFire alternatives.
826 * config/m68k/m68k.h (CLZ_DEFINED_VALUE_AT_ZERO): New macro.
827 * config/m68k/m68k.md (clzsi2): Define for ColdFire
828 architectures that support ff1 instruction.
830 2007-01-19 Richard Sandiford <richard@codesourcery.com>
831 Julian Brown <julian@codesourcery.com>
833 * config/m68k/m68k.h (CONST_OK_FOR_LETTER_P): Add an 'R' case.
834 * config/m68k/m68k.md (*movsi_cfv4): Fold into...
835 (*movsi_cf): ...here. Remove unnecessary 'R' from 'Rg'.
838 2007-01-19 Richard Sandiford <richard@codesourcery.com>
840 * config/m68k/m68k-protos.h (valid_mov3q_const): Take a HOST_WIDE_INT
842 (output_move_const_into_data_reg, output_move_simode_const): Delete.
843 * config/m68k/m68k.c (const_method, const_int_cost): Take a
844 HOST_WIDE_INT instead of an rtx.
845 (m68k_rtx_costs): Update call accordingly.
846 (output_move_const_into_data_reg): Likewise. Fix formatting.
847 (valid_mov3q_const): Take a HOST_WIDE_INT instead of an rtx.
849 (output_move_simode_const): Update calls after above changes.
850 Rework to use automatic variables and predicates like MEM_P.
851 * config/m68k/m68k.md (pushexthisi_const): Update call to
854 2007-01-19 Dirk Mueller <dmueller@suse.de>
856 * tree-ssa-alias.c (perform_var_substitution): Fix typo
859 2007-01-19 Richard Guenther <rguenther@suse.de>
861 * builtins.c (expand_builtin_cexpi): Fall back to expanding
862 via cexp in case sincos is not available.
864 2007-01-19 Richard Guenther <rguenther@suse.de>
866 * doc/tm.texi (TARGET_HAS_SINCOS): Document new target macro.
867 * defaults.h (TARGET_HAS_SINCOS): Default to off.
868 * config/linux.h (TARGET_HAS_SINCOS): Set to on if we have glibc.
869 * config/alpha/linux.h (TARGET_HAS_SINCOS): Likewise.
870 * config/sparc/linux.h (TARGET_HAS_SINCOS): Likewise.
871 * config/sparc/linux64.h (TARGET_HAS_SINCOS): Likewise.
872 * config/rs6000/linux.h (TARGET_HAS_SINCOS): Likewise.
873 * config/rs6000/linux64.h (TARGET_HAS_SINCOS): Likewise.
875 2007-01-19 Uros Bizjak <ubizjak@gmail.com>
877 * config/i386/i386.md (*fpatanxf3_i387, fpatan_extend<mode>xf3_i387):
879 (atan2sf3_1, atan2df3_1, atan2xf3_1): Remove insn patterns.
880 (atan2xf3): Directly generate RTL pattern.
881 (atan2<mode>3): Rename from atan2sf3 and atan2df3 and macroize insn
882 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
883 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
884 disable patterns for SSE math.
885 (atan<mode>2): Rename from atansf2 and atandf2 and macroize insn
886 patterns using X87MODEF12 mode macro. Use fpatan_extend<mode>xf3_i387
887 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
888 disable patterns for SSE math.
890 2007-01-19 Alexandre Oliva <aoliva@redhat.com>
892 * libgcc-std.ver: Fix typo in %inherit for GCC_4.3.0.
894 2007-01-18 Roger Sayle <roger@eyesopen.com>
896 * fold-const.c (fold_unary) <VIEW_CONVERT_EXPR>: Optimize away a
897 VIEW_CONVERT_EXPR to the same type as it's operand.
899 2007-01-18 David Edelsohn <edelsohn@gnu.org>
901 * config/rs6000/darwin-ldouble.c: Only build _SOFT_FLOAT if
902 configured for long double 128.
904 2007-01-18 Mike Stump <mrs@apple.com>
906 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Fix build
909 2007-01-18 Michael Meissner <michael.meissner@amd.com>
911 * i386.c (ix86_compute_frame_layout): Make fprintf's in #if 0 code
914 2007-01-18 Jan Hubicka <jh@suse.cz>
916 * tree-ssa-operands.c (vop_free_bucket_size): Never return value
917 greater than NUM_VOP_FREE_BUCKETS.
919 2007-01-18 Daniel Berlin <dberlin@dberlin.org>
921 * tree-ssa-structalias.c: Update comments.
922 (ptabitmap_obstack): Removed.
924 (oldpta_obstack): Ditto.
925 (stats): Add a few members.
926 (struct variable_info): Remove node, complex, address_taken, and
927 indirect_target members. Add oldsolution member.
928 (new_var_info): Do not initialize removed members.
929 (constraint_expr_type): Remove INCLUDES.
930 (constraint_graph): Add size, implicit_preds, rep,
931 indirect_cycles, eq_rep, label, direct_nodes, and complex members.
932 (FIRST_REF_NODE): New macro.
933 (LAST_REF_NODE): Ditto.
934 (FIRST_ADDR_NODE): Ditto.
935 (find): New function.
937 (dump_constraint): Do not handle INCLUDES.
938 (insert_into_complex): Do not insert duplicate constraints.
939 (condense_varmap_nodes): Renamed and rewritten into ...
940 (merge_node_constraints): This. Also fix bug in handling of
941 offseted copy constraints.
942 (clear_edges_for_node): No longer need to deal with preds at all,
943 or removing associated preds/succs.
944 (merge_graph_nodes): Deal with indirect_cycles.
945 Don't deal with predecessors.
946 (add_implicit_graph_edge): New function.
947 (add_pred_graph_edge): Ditto.
948 (add_graph_edge): Don't deal with predecessors.
949 (build_constraint_graph): Removed.
950 (build_pred_graph): New function.
951 (build_succ_graph): Ditto.
952 (struct scc_info): Removed in_component. Added roots, dfs, and
953 node_mapping. Remove visited_index, unification_queue.
954 (scc_visit): Deal with union-find we do now.
955 Deal with cycles with REF nodes.
956 (collapse_nodes): Renamed and rewritten to ...
958 (process_unification_queue): Removed.
959 (topo_visit): Cleanup
960 (do_da_constraint): Use find.
961 (do_sd_constraint): Ditto.
962 (do_ds_constraint): Ditto.
963 (do_complex_constraint): Ditto.
964 (init_scc_info): Update for removed and added members.
965 (find_and_collapse_graph_cycles): Renamed and rewritten into ...
966 (find_indirect_cycles): This.
967 (equivalence_class): New variable.
968 (label_visit): New function.
969 (perform_variable_substitution): Rewritten.
970 (free_var_substitution_info): New function.
971 (find_equivalent_node): Ditto.
972 (move_complex_constraints): Ditto.
973 (eliminate_indirect_cycles): Ditto.
974 (solve_graph): Only propagate changed bits.
975 Use indirect cycle elimination.
977 (tree_id_t): Rename to tree_vi_t, delete id member, add vi member.
978 (tree_id_eq): Renamed to ...
979 (tree_vi_eq): This. Update for member change
980 (insert_id_for_tree): Renamed and rewritten to ...
981 (insert_vi_for_tree): This.
982 (lookup_id_for_tree): Renamed and rewritten to ...
983 (lookup_vi_for_tree): This.
984 (get_id_for_tree): Renamed and rewritten to ...
985 (get_vi_for_tree): Ditto.
986 (get_constraint_exp_from_ssa_var): Update to use get_vi_for_tree.
987 (process_constraint): Don't handle INCLUDES.
988 Remove special ADDRESSOF case.
989 (find_func_aliases): Rewrite to use vi functions instead of id
991 (create_function_info_for): Ditto.
992 (create_variable_info_for): Ditto.
993 (intra_create_variable_infos): Ditto.
994 (merge_smts_into): Ditto.
995 (find_what_p_points_to): Ditto.
996 (init_base_vars): Ditto.
997 (init_alias_vars): Ditto.
998 (remove_preds_and_fake_succs): New function.
999 (dump_sa_points_to_info): Dump new stats.
1000 (dump_solution_for_var): Use find.
1001 (set_used_smts): Fix formatting.
1002 (compute_points_to_sets): Updated for new functions.
1003 (ipa_pta_execute): Ditto.
1005 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1006 Richard Sandiford <richard@codesourcery.com>
1008 * doc/tm.texi (TARGET_FUNCTION_VALUE): Expand documentation of
1010 * calls.c (expand_call): If the return value is a PARALLEL,
1011 extract its first member.
1012 * config/m68k/linux.h (FUNCTION_EXTRA_EPILOGUE): Remove.
1013 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1014 use FUNCTION_EXTRA_EPILOGUE.
1015 (m68k_function_value): Return a PARALLEL if the return value
1016 is of a pointer type.
1017 * config/m68k/netbsd-elf.h (current_function_returns_pointer)
1018 (FUNCTION_EXTRA_EPILOGUE): Remove.
1019 * config/m68k/m68k.md (D0_REG): New constant.
1021 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1023 * config/m68k/m68k.c (m68k_output_function_epilogue): Don't
1024 output a NOP for empty epilogues.
1026 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1028 * config/m68k/m68k.c (m68k_use_return_insn): Update comments
1029 before function. Extend register save check to include all
1030 registers, not just integer ones.
1032 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1034 * config/m68k/m68k-protos.h (use_return_insn): Rename to...
1035 (m68k_use_return_insn): ...this.
1036 * config/m68k/m68k.h (USE_RETURN_INSN): Delete.
1037 * config/m68k/m68k.c (use_return_insn): Rename to...
1038 (m68k_use_return_insn): ...this.
1039 * config/m68k/m68k.md (return): Use m68k_use_return_insn instead
1042 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1044 * config/m68k/fpgnulib.c (__truncdfsf2): Implement round to
1045 nearest even, fix denormal rounding overflow.
1047 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1049 * config/m68k/m68k.md (movsf_cf_hard): Use fsmove instead of
1050 f%$smove and f%$move.
1051 (movdf_cf_hard): Use fdmove for cases 0 and 3 and fmove for case 1.
1052 (extendsfdf2_cf): Use fdmove instead of f%&move.
1053 (truncdfsf2_cf): Use fsmove instead of f%$smove.
1054 (add<mode>3_cf, sub<mode>3_cf): Use <FP:prec> instead of <FP:round>.
1056 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1058 * config/m68k/m68k.md (movdf_cf_hard): Use output_move_double for
1061 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1063 * real.h (real_format): Add a canonical_nan_lsbs_set field.
1064 (coldfire_single_format): Declare.
1065 (coldfire_double_format): Likewise.
1066 * real.c (encode_ieee_single): Use canonical_nan_lsbs_set instead
1067 of qnan_msb_set to determine the lower bits of a canonical
1069 (encode_ieee_double): Likewise.
1070 (encode_ieee_quad): Likewise.
1071 (ieee_single_format): Initialize canonical_nan_lsbs_set.
1072 (mips_single_format): Likewise.
1073 (ieee_double_format): Likewise.
1074 (mips_double_format): Likewise.
1075 (ieee_extended_motorola_format): Likewise.
1076 (ieee_extended_intel_96_format): Likewise.
1077 (ieee_extended_intel_128_format): Likewise.
1078 (ieee_extended_intel_96_round_53_format): Likewise.
1079 (ibm_extended_format): Likewise.
1080 (mips_extended_format): Likewise.
1081 (ieee_quad_format): Likewise.
1082 (mips_quad_format): Likewise.
1083 (vax_f_format): Likewise.
1084 (vax_d_format): Likewise.
1085 (vax_g_format): Likewise.
1086 (i370_single_format): Likewise.
1087 (i370_double_format): Likewise.
1088 (decimal_single_format): Likewise.
1089 (decimal_double_format): Likewise.
1090 (decimal_quad_format): Likewise.
1091 (c4x_single_format): Likewise.
1092 (c4x_extended_format): Likewise.
1093 (real_internal_format): Likewise.
1094 (coldfire_single_format): New real_format.
1095 (coldfire_double_format): Likewise.
1096 * config/pdp11/pdp11.c (pdp11_f_format): Initialize
1097 canonical_nan_lsbs_set.
1098 (pdp11_d_format): Likewise.
1099 * config/m68k/m68k.c (override_options): Override REAL_FORMAT_MODE
1100 if TARGET_COLDFIRE_CPU.
1102 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1104 * config/m68k/m68k-protos.h (m68k_output_pic_call): Delete.
1105 (output_call, m68k_legitimize_call_address): Declare.
1106 * config/m68k/m68k.h (EXTRA_CONSTRAINT): Remove unnecessary
1107 parenthesees. Add support for a 'W' constraint.
1108 (LEGITIMATE_PIC_OPERAND_P): Remove SYMBOL_REF_FLAG handling.
1109 (PRINT_OPERAND_PUNCT_VALID_P): Remove comment about 'o'.
1110 (m68k_symbolic_call, m68k_symbolic_jump): Declare.
1111 * config/m68k/m68k.c (m68k_symbolic_call, m68k_symbolic_jump): New
1113 (override_options): Initialize them. Do not set flag_no_function_cse
1114 for TARGET_ID_SHARED_LIBRARY.
1115 (m68k_output_pic_call): Delete.
1116 (m68k_legitimize_call_address): New function.
1117 (print_operand): Remove the %o prefix. Handle the %p prefix.
1118 (output_call): New function.
1119 (m68k_output_mi_thunk): Use m68k_symbolic_jump. Always load the
1120 target address from the GOT if symbolic jumps are not allowed.
1121 * config/m68k/m68k.md (call, general_operand): Do not set
1122 SYMBOL_REF_FLAG. Use m68k_legitimize_call_address instead.
1123 Merge separate flag_pic and !flag_pic define_insns into...
1124 (*call, *call_value): ...these new patterns. Match the address
1125 rather than the containing MEM and require it to be a call_operand.
1126 Use output_call to generate the asm template.
1127 * config/m68k/predicates.md (const_call_operand): New predicate.
1128 (call_operand): Likewise.
1130 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1132 * config/m68k/m68k.h (REGISTER_MOVE_COST): Simplify definition.
1133 (STACK_GROWS_DOWNWARD): Define to 1.
1134 (FUNCTION_VALUE, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P): Equivocate
1135 comments, emphasizing that these values are only defaults.
1136 * config/m68k/linux.h (LINK_SPEC): Fix formatting in #undef.
1137 * config/m68k/m68k.c (const_method): Remove trailing whitespace.
1139 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1141 * config/m68k/m68k.md (cmpsi): Remove outdated flag_pic handling.
1143 2007-01-18 Kazu Hirata <kazu@codesourcery.com>
1144 Richard Sandiford <richard@codesourcery.com>
1146 * config/m68k/m68k.h (DATA_REGNO_P, ADDRESS_REGNO_P, INT_REGNO_P)
1147 (FP_REGNO_P): New macros.
1148 (REGNO_OK_FOR_INDEX_P, REGNO_OK_FOR_BASE_P, REGNO_OK_FOR_DATA_P)
1149 (REGNO_OK_FOR_FP_P, REG_OK_FOR_INDEX_P, REG_OK_FOR_BASE_P): Use them.
1150 (EH_RETURN_STACKADJ_RTX): Use A0_REG.
1151 * config/m68k/m68k.c (m68k_regno_mode_ok): Use the new REGNO macros.
1153 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1155 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1156 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-rtems*): Add t-floatlib
1158 * config/m68k/t-floatlib: New file, extracting common code from...
1159 * config/m68k/t-m68kbare, config/m68k/t-m68kelf,
1160 * config/m68k/t-uclinux: Here.
1161 * config/m68k/fpgnulib.c: Do not compile extendeed precision
1162 routines on ColdFire targets.
1164 2007-01-18 Nathan Sidwell <nathan@codesourcery.com>
1166 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1167 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*)
1168 (m68k-*-uclinuxoldabi, m68k-*-uclinux*, m68k-*-linux*)
1169 (m68k-*-rtems*): Use tm_file rather than m68k/m68k.h and
1170 explicitly set MOTOROLA to 1.
1171 * config/m68k/m68k.h (MOTOROLA): Simplify definition accordingly.
1173 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1174 Nathan Sidwell <nathan@codesourcery.com>
1176 * config/m68k/m68k.h (PCC_STATIC_STRUCT_RETURN): Delete.
1177 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Add MOTOROLA cases.
1178 * config/m68k/coff.h (REGISTER_PREFIX_MD): Delete.
1179 * config/m68k/m68020-elf.h (LIB_SPEC): Delete.
1180 * config/m68k/m68k-none.h (CC1_SPEC, CPP_SUBTARGET_SPEC): Delete.
1181 * config/m68k/m68kelf.h (IMMEDIATE_PREFIX, REGISTER_PREFIX_MD)
1182 (ASM_OUTPUT_REG_PUSH, ASM_OUTPUT_REG_POP): Delete.
1183 (NO_DOLLAR_IN_LABEL, PCC_STATIC_STRUCT_RETURN): Don't undefine.
1184 (BSS_ASM_OP, ASM_OUTPUT_SKIP, ASM_OUTPUT_ASCII): Delete.
1185 * config/m68k/m68kemb.h (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1186 * config/m68k/linux.h (SIZE_TYPE, PTRDIFF_TYPE, WCHAR_TYPE)
1187 (WCHAR_TYPE_SIZE, TARGET_OBJFMT_CPP_BUILTINS): Delete.
1188 (TARGET_OS_CPP_BUILTINS): Don't define mc68000 and mc68020 here.
1189 (DBX_REGISTER_NUMBER): Delete.
1190 * config/m68k/netbsd-elf.h (IMMEDIATE_PREFIX): Delete.
1191 (PCC_STATIC_STRUCT_RETURN): Don't undefine.
1192 * config/m68k/openbsd.h (PCC_STATIC_STRUCT_RETURN): Define.
1194 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1196 * config.gcc (m68k-*-uclinux*): Add flat.h to $tm_file.
1197 * config/flat.h: New file.
1198 * crtstuff.c (USE_PT_GNU_EH_FRAME): Don't define if
1200 * config/m68k/m68k.h (ASM_PREFERRED_EH_DATA_FORMAT): Do not use
1201 indirect references for -msep-data or -mid-shared-library.
1202 Do not use PC-relative code addresses either.
1203 * config/m68k/m68k.c (override_options): Restrict -fPIC error
1205 * config/m68k/uclinux.h (STARTFILE_SPEC): Define. Use Scrt1.o
1206 for shared libraries and crt1.o for executables. Use crti.o and
1208 (ENDFILE_SPEC): Use crtend.o and crtn.o.
1209 (LIB_SPEC): Suppress -Rlibc.gdb if -static-libc is given.
1210 Do not add -elf2flt or -shared-lib-id options here.
1211 (LINK_SPEC): Define. Pass -elf2flt if no -elf2flt option is given.
1212 Pass -shared-lib-id if -mid-shared-library, taking the library
1213 identifier from -mshared-library-id if given, otherwise
1215 (EH_FRAME_IN_DATA_SECTION): Do not undefine.
1216 (INIT_SECTION_ASM_OP, FINI_SECTION_ASM_OP): Likewise.
1217 (TARGET_OS_CPP_BUILTINS): Define __GXX_MERGED_TYPEINFO_NAMES=0
1218 and __GXX_TYPEINFO_EQUALITY_INLINE=0 if -mid-shared-library.
1219 (DRIVER_SELF_SPECS): Map unadorned PIC options to -msep-data.
1220 * config/m68k/t-uclinux (EXTRA_MULTILIB_PARTS): Add crtbegin.o
1222 * config/m68k/lb1sf68.asm (PICCALL): Use an lea and pc-relative
1223 jump sequence for ISA A and ISA A+.
1224 (PICJUMP): Likewise.
1226 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1228 * config.gcc (m68k-*-uclinux*): Base the port on the common
1229 and m68k GNU/Linux files rather than on the generic ELF ones.
1230 * config/m68k/uclinux.h (TARGET_VERSION): Override.
1231 (TARGET_OS_CPP_BUILTINS): Use LINUX_TARGET_OS_CPP_BUILTINS.
1233 2007-01-18 Julian Brown <julian@codesourcery.com>
1234 Richard Sandiford <richard@codesourcery.com>
1236 * config/m68k/m68k.h (LONG_DOUBLE_TYPE_SIZE): Make 64-bit on ColdFire.
1237 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1238 * config/m68k/netbsd-elf.h (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Undefine
1240 * config/m68k/uclinux-oldabi.h (LONG_DOUBLE_TYPE_SIZE): Redefine to
1242 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1244 2007-01-18 Richard Sandiford <richard@codesourcery.com>
1246 * doc/install.texi: Document m68k-uclinuxoldabi.
1247 * config.gcc (m68k-*-uclinuxoldabi): New configuration.
1248 * config/m68k/uclinux-oldabi.h: New file, copied from
1249 config/m68k/uclinux.h.
1251 2007-01-18 Kaz Kojima <kkojima@rr.iij4u.or.jp>
1253 * config/m32r/m32r-protos.h (m32r_expand_epilogue): Declare it.
1254 * config/m32r/m32r.c (m32r_setup_incoming_varargs): Use gen_frame_mem.
1255 (m32r_compute_frame_size): Use unsigned for regno.
1256 (m32r_reload_lr): Use gen_frame_mem.
1258 (m32r_output_function_epilogue): Don't output the function epilogue
1260 (m32r_expand_epilogue): New.
1261 (direct_return): Return false if the function has the interrupt
1263 (m32r_hard_regno_rename_ok): Remove code for the textual epilogue.
1264 * config/m32r/m32r.md (epilogue): New expander.
1265 (return_lr, return_rte): New insns.
1266 (return): Make it expander.
1267 (return_normal): New expander.
1269 2007-01-18 Josh Conner <jconner@apple.com>
1272 * config/rs6000/rs6000.c (rs6000_emit_vector_compare): Add
1273 support for UNLE, UNLT, UNGE, and UNGT.
1275 2007-01-18 Jan Hubicka <jh@suse.cz>
1277 * tree-vrp.c (finalize_jump_threads): Do not call cleanup_cfg by hand.
1278 * tree-tailcall (add_virtual_phis): Likewise.
1279 (optimize_tail_call): Return TODOs.
1280 (execute_tail_calls): Return TODOs.
1281 * tree-ssa-ccp (execute_fold_all_builtins): Do cleanup_cfg via TODO.
1282 * tree-cfgcleanup.c (cleanup_tree_cfg_loop): Return if something
1284 * tree-ssa-forwprop.c (tree_ssa_forward_propagate_single_use_value):
1285 Cleanup cfg using TODO.
1286 * tree-flow.h (cleanup_tree_cfg_loop): Update prototype.
1287 * passes.c (execute_function_todo): When cleanup did something, remove
1289 * tree-cfg.c (pass_build_cfg): Add cleanup_cfg TODO.
1290 (make_edges): Don't cleanup_cfg.
1292 2007-01-18 Uros Bizjak <ubizjak@gmail.com>
1294 * reg-stack.c (subst_stack_regs_pat) [UNSPEC_SINCOS_COS,
1295 UNSPEC_XTRACT_FRACT]: Use generic code for instructions that
1296 operate on the top of stack.
1297 [UNSPEC_SINCOS_SIN, UNSPEC_XTRACT_EXP, UNSPEC_TAN]: Rewrite
1298 register handling of instructions that output to the second
1300 [UNSPEC_TAN_ONE, UNSPEC_TAN_TAN]: Remove.
1301 (move_for_stack_reg): Special-case check for dead destination
1302 stack slot for constant load of 1.0 inside UNSPEC_TAN.
1304 * config/i386/i386.md (UNSPEC_TAN): New constant.
1305 (UNSPEC_TAN_ONE, UNSPEC_TAN_TAN): Remove.
1306 (fptanxf4_i387, fptan_extend<mode>xf4_i387): New patterns
1307 to correctly model move of constant 1.0 to top stack slot.
1308 (*tandf3_1, *tansf3_1, *tanxf3_1): Remove insn patterns.
1309 (unnamed peephole2 pattern): Remove corresponding peephole2
1310 pattern that optimizes tan insn and loading of constant 1.0.
1311 (tanxf2): Use fptanxf4_i387.
1312 (tan<mode>2): Rename from tansf2 and tandf2 and macroize insn
1313 patterns using X87MODEF12 mode macro. Use fptan_extend<mode>xf4_i387
1314 and truncate result to requested mode. Use SSE_FLOAT_MODE_P to
1315 disable patterns for SSE math.
1316 (sincos<mode>3): Use truncxf<mode>2_i387_noop for truncation.
1317 (fyl2x_extend<mode>xf3_i387): Use X87MODEF12 for operand 1.
1319 2007-01-18 Dirk Mueller <dmueller@suse.de>
1320 Richard Guenther <rguenther@suse.de>
1323 * doc/invoke.texi (Warray-bounds): Document -Warray-bounds.
1324 * common.opt (Warray-bounds): Add new warning option.
1325 * c-opts.c (c_common_handle_option): Define -Warray-bounds
1327 * Makefile.in: make tree-vrp.o depend on toplev.h
1328 * tree-vrp.c (vrp_finalize): Call check_array_refs if -Warray-bounds
1330 (check_array_refs, check_array_bounds, check_array_ref): New.
1332 2007-01-18 Jan Hubicka <jh@suse.cz>
1334 * tree-ssa-ccp.c (ccp_finalize): Return if something changed.
1335 (execute_ssa_ccp): Return flags conditionally.
1336 * tree-ssa-propagate.c (substitue_and_fold): Return if something was
1338 * tree-ssa-propagate.h (substitute_and_fold): Update prototype.
1340 2007-01-18 Steven Bosscher <steven@gcc.gnu.org>
1342 * cfgcleanup.c (cleanup_cfg): Detect cfglayout mode and set
1343 the CLEANUP_CFGLAYOUT flag when in cfglayout mode.
1345 * Makefile.c (GTFILES): Add cfglayout.h.
1346 * gengtype.c (open_base_files): Likewise.
1347 * cfglayout.c (cfg_layout_function_footer,
1348 cfg_layout_function_header) Reindent to make gengtype happy.
1349 * cfglayout.h (cfg_layout_function_footer,
1350 cfg_layout_function_header): Add GTY(()) marker.
1352 * ifcvt.c (noce_try_sign_mask): Make sure INSN_B is non-null.
1354 2007-01-18 Ben Elliston <bje@au.ibm.com>
1356 * genautomata.c (write_automata): Include xstrerror output in the
1357 error message if writing the DFA description file fails.
1359 2007-01-17 H.J. Lu <hongjiu.lu@intel.com>
1361 * config/mips/mips-protos.h (mips_output_external): Make it
1363 * config/mips/iris.h (TARGET_ASM_EXTERNAL_LIBCALL): Removed.
1364 * config/mips/mips.c (irix_output_external_libcall): Likewise.
1365 (extern_list): Likewise.
1366 (extern_head): Likewise.
1367 (TARGET_ASM_FILE_END): Likewise.
1368 (mips_file_end): Likewise.
1369 (mips_output_external): Rewritten.
1371 2007-01-18 Ben Elliston <bje@au.ibm.com>
1373 * genpreds.c (write_insn_preds_c): Only write out the function
1374 body for regclass_for_constraint if we have register constraints.
1376 2007-01-17 Tom Tromey <tromey@redhat.com>
1378 * doc/sourcebuild.texi (libgcj Tests): Use sourceware.org.
1379 * doc/install.texi (Testing): Use sourceware.org.
1380 (Binaries): Likewise.
1381 (Specific): Likewise.
1382 * doc/contrib.texi (Contributors): Use sourceware.org.
1384 2007-01-17 Anatoly Sokolov <aesok@post.ru>
1386 * config/avr/avr.h (AVR_HAVE_LPMX): New macro.
1387 (AVR_ENHANCED): Rename to ...
1388 (AVR_HAVE_MUL): ... new.
1389 (avr_enhanced_p): Rename to ...
1390 (avr_have_mul_p): ... new.
1391 (TARGET_CPU_CPP_BUILTINS): Use 'avr_have_mul_p' instead of
1392 'avr_enhanced_p' for "__AVR_ENHANCED__". Define "__AVR_HAVE_MUL__".
1393 * config/avr/avr.c (avr_enhanced_p): Rename to ...
1394 (avr_have_mul_p): ... new.
1395 (base_arch_s): Rename 'enhanced' to 'have_mul'.
1396 (avr_override_options): Use 'avr_have_mul_p' and 'have_mul' instead of
1397 'avr_enhanced_p' and 'enhanced'.
1398 (ashlhi3_out, ashrhi3_out, lshrhi3_out, avr_rtx_costs): Use
1399 AVR_HAVE_MUL instead of AVR_ENHANCED.
1400 * avr.md (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1401 (mulqi3, *mulqi3_enh, *mulqi3_call, mulqihi3, umulqihi3, mulhi3,
1402 *mulhi3_enh, *mulhi3_call, mulsi3, *mulsi3_call): Use AVR_HAVE_MUL
1403 instead of AVR_ENHANCED.
1404 (*tablejump_enh): Use AVR_HAVE_LPMX instead of AVR_ENHANCED.
1405 * libgcc.S: Use __AVR_HAVE_MUL__ instead of __AVR_ENHANCED__.
1406 (__tablejump__): Use __AVR_HAVE_LPMX__ instead of __AVR_ENHANCED__.
1408 2007-01-17 Ian Lance Taylor <iant@google.com>
1410 * vec.h (VEC_reserve_exact): Define.
1411 (vec_gc_p_reserve_exact): Declare.
1412 (vec_gc_o_reserve_exact): Declare.
1413 (vec_heap_p_reserve_exact): Declare.
1414 (vec_heap_o_reserve_exact): Declare.
1415 (VEC_OP (T,A,reserve_exact)): New static inline function, three
1417 (VEC_OP (T,A,reserve)) [all versions]: Remove handling of
1419 (VEC_OP (T,A,alloc)) [all versions]: Call ...reserve_exact.
1420 (VEC_OP (T,A,copy)) [all versions]: Likewise.
1421 (VEC_OP (T,a,safe_grow)) [all versions]: Likewise.
1422 * vec.c (calculate_allocation): Add exact parameter. Change all
1424 (vec_gc_o_reserve_1): New static function, from vec_gc_o_reserve.
1425 (vec_gc_p_reserve, vec_gc_o_reserve): Call vec_gc_o_reserve_1.
1426 (vec_gc_p_reserve_exact, vec_gc_o_reserve_exact): New functions.
1427 (vec_heap_o_reserve_1): New static function, from vec_heap_o_reserve.
1428 (vec_heap_p_reserve, vec_heap_o_reserve): Call vec_heap_o_reserve_1.
1429 (vec_heap_p_reserve_exact): New function.
1430 (vec_heap_o_reserve_exact): New function.
1432 2007-01-17 Jan Hubicka <jh@suse.cz>
1434 * ipa-type-escape.c (look_for_casts): Revamp using handled_component_p.
1436 2007-01-17 Eric Christopher <echristo@apple.com>
1438 * config.gcc: Support core2 processor.
1440 2007-01-16 Jan Hubicka <jh@suse.cz>
1442 * tree-ssanames.c (release_dead_ssa_names): Instead of ggc_freeing
1443 the names, just unlink the chain so we don't crash on dangling pointers
1446 2007-01-16 Jan Hubicka <jh@suse.cz>
1448 * cgraph.h (cgraph_decide_inlining_incrementally): Kill.
1449 * tree-pass.h: Reorder to make IPA passes appear toegher.
1450 (pass_early_inline, pass_inline_parameters, pass_apply_inline): Declare.
1451 * cgraphunit.c (cgraph_finalize_function): Do not compute inling
1452 parameters, do not call early inliner.
1453 * ipa-inline.c: Update comments. Include tree-flow.h
1454 (cgraph_decide_inlining): Do not compute inlining parameters.
1455 (cgraph_decide_inlining_incrementally): Return TODOs; assume to
1456 be called with function context set up.
1457 (pass_ipa_inline): Remove unreachable functions before pass.
1458 (cgraph_early_inlining): Simplify assuming to be called from the
1460 (pass_early_inline): New pass.
1461 (cgraph_gate_ipa_early_inlining): New gate.
1462 (pass_ipa_early_inline): Turn into simple wrapper.
1463 (compute_inline_parameters): New function.
1464 (gate_inline_passes): New gate.
1465 (pass_inline_parameters): New pass.
1466 (apply_inline): Move here from tree-optimize.c
1467 (pass_apply_inline): New pass.
1468 * ipa.c (cgraph_remove_unreachable_nodes): Verify cgraph after
1470 * tree-inline.c (optimize_inline_calls): Return TODOs rather than
1472 (tree_function_versioning): Do not allocate dummy struct function.
1473 * tree-inline.h (optimize_inline_calls): Update prototype.
1474 * tree-optimize.c (execute_fixup_cfg): Export.
1475 (pass_fixup_cfg): Remove
1476 (tree_rest_of_compilation): Do not apply inlines.
1477 * tree-flow.h (execute_fixup_cfg): Declare.
1478 * Makefile.in (gt-passes.c): New.
1479 * passes.c: Include gt-passes.h
1480 (init_optimization_passes): New passes.
1481 (nnodes, order): New static vars.
1482 (do_per_function_toporder): New function.
1483 (execute_one_pass): Dump current pass here.
1484 (execute_ipa_pass_list): Don't dump current pass here.
1486 2007-01-16 Janis Johnson <janis187@us.ibm.com>
1488 * config/dfp-bit.c (dfp_compare_op): Return separate value for NaN.
1489 (DFP_NE, DFP_LE, DFP_GE): Return false for NaN.
1491 2007-01-16 David Edelsohn <edelsohn@gnu.org>
1493 * config/rs6000/darwin-ldouble.c: Build file for SOFT_FLOAT.
1494 (strong_alias): Define.
1495 (__gcc_qmul): Provide non-FMA for soft-float.
1501 (__gcc_qunord): New.
1511 * config/rs6000/rs6000.c (rs6000_init_libfuncs): Initialize
1512 soft-float functions.
1513 * config/rs6000/libgcc-ppc-glibc.ver: Version soft-float symbols.
1514 * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Do not warn
1515 about long double soft float.
1517 2007-01-16 Dorit Nuzman <dorit@il.ibm.com>
1518 Tehila Meyzels <tehila@il.ibm.com>
1520 * tree-vectorizer.h (is_pattern_stmt_p): New.
1521 * tree-vect-analyze.c (vect_determine_vectorization_factor): Fix
1522 formatting (tabs instead of spaces). Cleanup and clarify setting
1523 of STMT_VINFO_VECTYPE. Call is_pattern_stmt_p.
1524 * tree-vect-transform.c (vect_get_vec_def_for_operand): Fix typo.
1525 (vectorizable_type_demotion): Check that types are integral.
1526 (vectorizable_type_promotion): Likewise.
1527 (vectorizable_store): Fix typo. Eliminate new-line at end of
1530 2007-01-16 Jan Hubicka <jh@suse.cz>
1532 * tree-ssanames.c (release_dead_ssa_names): Remove invalidated
1535 2007-01-15 Eric Christopher <echristo@apple.com>
1537 * ifcvt.c: Include vec.h, vecprim.h.
1538 (check_cond_move_block): New argument regs.
1539 Reorganize. Add registers used to regs.
1540 (cond_move_process_if_block): Use regs set above as
1543 2007-01-15 Eric Christopher <echristo@apple.com>
1545 * config/darwin.h: Update copyright.
1546 (TARGET_OPTION_TRANSLATE_TABLE): Add umbrella.
1547 (LINK_COMMAND_SPEC): Add -u.
1548 (LINK_SPEC): Fix umbrella for above.
1550 2007-01-15 Joseph S. Myers <joseph@codesourcery.com>
1552 * config/soft-fp/op-common.h, config/soft-fp/op-4.h: Update from
1555 2007-01-15 Tom Tromey <tromey@redhat.com>
1557 * doc/sourcebuild.texi (libgcj Tests): Don't mention jacks.
1558 * doc/install.texi (Testing): Don't mention jacks.
1559 (Configuration): Document --enable-java-maintainer-mode. Move
1560 --with-java-home to libgcj-specific section. Document
1562 (Prerequisites): Mention --enable-java-maintainer-mode, ecj1.
1564 2007-01-15 Jan Hubicka <jh@suse.cz>
1566 * tree-ssa-dce.c (DCE_TODOs): New.
1567 (propagate_necessity): Return if something changed.
1568 (eliminate_unnecessary_stmts): Likewise.
1569 (perform_tree_ssa_dce): Return TODO flags when needed.
1570 (pass_dce, pass_dce_loop, pass_cd_dce): Remove TODO flags.
1572 2007-01-15 Uros Bizjak <ubizjak@gmail.com>
1574 * config/i386/i386.md (fyl2xxf3_i387): Rename from fyl2x_xf3.
1575 (fyl2x_extend<mode>xf3_i387): New insn pattern.
1576 (log<mode>2): Rename from logsf2 and logdf2 and macroize insn
1577 patterns using X87MODEF12 mode macro. Extend operand 1
1578 to XFmode. Use SSE_FLOAT_MODE_P to disable patterns for SSE math.
1579 (log10<mode>2): Ditto.
1580 (log2<mode>2): Ditto.
1581 (log1p<mode>2): Ditto.
1582 (logb<mode>2): Ditto.
1583 (fyl2xp1xf3_i387): Rename from fyl2xp1_xf3.
1584 (fyl2xp1_extend<mode>xf3_i387): New insn pattern.
1585 (*fxtractxf3_i387): Rename from *fxtractxf3.
1586 (fxtract_extend<mode>xf3_i387): New insn pattern.
1587 (ilogbsi2): Use match_dup 3, not match_operand:XF 3.
1588 * config/i386/i386.c (ix86_emit_i387_log1p): Use gen_fyl2xp1xf3_i387()
1589 and gen_fyl2xxf3_i387().
1591 2007-01-14 Zdenek Dvorak <dvorakz@suse.cz>
1593 * loop-unswitch.c (unswitch_loop): Do not call fix_loop_placement.
1594 * cfgloopmanip.c (fix_loop_placement): Made static. Use
1595 get_loop_exit_edges. Changed return type to bool.
1596 * cfgloop.h (fix_loop_placement): Declaration removed.
1598 2007-01-14 Dorit Nuzman <dorit@il.ibm.com>
1600 * param.h (MIN_VECT_LOOP_BOUND): New.
1601 * params.def (MIN_VECT_LOOP_BOUND): New.
1602 * tree-vectorizer.c (slpeel_tree_peel_loop_to_edge): Takes another
1603 argument - minimum threshold for number of iterations.
1604 * tree-vectorizer.h (slpeel_tree_peel_loop_to_edge): Add another
1605 argument to declaration.
1606 * tree-vect-analyze.c (vect_analyze_operations): Check value of
1607 MIN_VECT_LOOP_BOUND.
1608 * tree-vect-transform.c (vect_do_peeling_for_loop_bound): Call
1609 slpeel_tree_peel_loop_to_edge with additional argument.
1610 (vect_do_peeling_for_alignment): Likewise.
1611 * doc/invoke.texi (min-vect-loop-bound): Document new param option.
1613 2007-01-14 Uros Bizjak <ubizjak@gmail.com>
1616 * config/i386/i386.c (print_operand) ['z']: Output 'b' for
1619 2007-01-14 Jan Hubicka <jh@suse.cz>
1621 * tree-dfa.c (remove_referenced_var): New function.
1622 * tree-ssa-live.c (remove_unused_locals): Walk referenced vars and
1623 prune referenced vars list too.
1624 * tree-flow.h (remove_referenced_var): Declare.
1626 2007-01-14 Jan Hubicka <jh@suse.cz>
1628 * tree-eh.c (add_stmt_to_eh_region_fn): Do not add call_exprs
1630 (remove_stmt_from_eh_region_fn): Do not remove call_exprs.
1631 (verify_eh_throw_stmt_node, verify_eh_throw_table_statements): Kill.
1632 * except.h (verify_eh_throw_table_statements): Kill prototype.
1633 * cfgexpand.c (expand_gimple_basic_block): Propagate Eh regions
1634 into call exrepssions.
1635 * tree-optimize.c (execute_free_cfg_annotatiosn): Do not call
1637 * tree-cfg.c: Include pointer-set.h.
1638 (verify_node_sharing): Work on pointer set.
1639 (verify_eh_throw_stmt_node): New.
1640 (verify_stmts): Use pointers sets, verify throw_stmt.
1642 2007-01-13 Zdenek Dvorak <dvorakz@suse.cz>
1644 * ipa-reference.c (analyze_function): Consider also addresses taken
1647 2007-01-12 Roger Sayle <roger@eyesopen.com>
1649 * c-typeck.c (null_pointer_constant_p): Replace use of
1650 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1651 (build_c_cast): Likewise.
1653 2007-01-12 Roger Sayle <roger@eyesopen.com>
1655 * tree.h (force_fit_type_double): Remove unused final argument.
1656 * c-common.c (constant_expression_warning): Replace use of
1657 TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
1658 (convert_and_check): Likewise.
1659 (shorten_compare): Update call to force_fit_type_double.
1660 (c_common_truthvalue_conversion) <INTEGER_CST>: Use integer_zerop.
1661 * convert.c (convert_to_pointer): Update call to
1662 force_fit_type_double.
1663 * fold-const.c (force_fit_type_double): Remove overflowed_const
1665 (int_const_binop, fold_convert_const_int_from_int,
1666 fold_convert_const_int_from_real, fold_div_compare,
1667 fold_sign_changed_comparison, fold_unary, fold_negate_const,
1668 fold_abs_const, fold_not_const): Remove the final argument from
1669 calls to force_fit_type_double.
1671 2007-01-12 Andrew Pinski <andrew_pinski@playstation.sony.com>
1673 * configure.ac: Set insn to "nop" for spu-*-* also.
1674 * configure: Regenerate.
1676 2007-01-12 Olga Golovanevsky <olga@il.ibm.com>
1678 * builtins.def : Add BUILT_IN_FREE.
1680 2007-01-12 Jan Hubicka <jh@suse.cz>
1682 PR tree-optimization/30443
1683 * tree-inline.c (tree_function_versioning): Do not optimize when
1684 cloning for inlining.
1686 2007-01-12 Zdenek Dvorak <dvorakz@suse.cz>
1688 * doc/loop.texi: Document recording of loop exits.
1689 * cfgloopmanip.c (loopify, duplicate_loop): Use alloc_loop.
1690 (update_single_exits_after_duplication,
1691 update_single_exit_for_duplicated_loop,
1692 update_single_exit_for_duplicated_loops): Removed.
1693 (duplicate_loop_to_header_edge): Do not call
1694 update_single_exits_after_duplication and
1695 update_single_exit_for_duplicated_loops.
1696 (loop_version): Do not update single_exit information.
1697 (fix_loop_structure): Use record_loop_exits instead of
1698 mark_single_exit_loops.
1699 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update
1700 the lists of loop exits.
1701 * cfghooks.c (redirect_edge_and_branch, redirect_edge_and_branch_force,
1702 split_edge, merge_blocks): Update the lists of loop exits.
1703 * modulo-sched.c (sms_schedule): Pass LOOPS_HAVE_RECORDED_EXITS to
1704 loop_optimizer_init.
1705 * loop-init.c (loop_optimizer_init): Call record_loop_exits instead
1706 of mark_single_exit_loops.
1707 (loop_optimizer_finalize): Call release_recorded_exits.
1708 * tree-ssa-loop.c (tree_loop_optimizer_init): Pass
1709 LOOPS_HAVE_RECORDED_EXITS to loop_optimizer_init.
1710 * tree-vectorizer.c (slpeel_tree_duplicate_loop_to_edge_cfg): Do not
1711 update single exit information.
1712 * lambda-code.c (perfect_nestify): Ditto.
1713 * cfgloop.c (flow_loop_free): Destroy the list of exits of the loop.
1714 (mark_single_exit_loops): Removed.
1715 (alloc_loop, loop_exit_hash, loop_exit_eq, loop_exit_free,
1716 get_exit_descriptions, rescan_loop_exit, record_loop_exits,
1717 dump_recorded_exit, dump_recorded_exits, release_recorded_exits): New
1719 (get_loop_exit_edges, single_exit): Use recorded exit lists.
1720 (add_bb_to_loop, remove_bb_from_loops): Update the lists of loop exits.
1721 (verify_loop_structure): Verify consistency of the exit lists.
1722 (flow_loops_find): Use alloc_loop. Initialize exits hash.
1723 (set_single_exit): Removed.
1724 * cfgloop.h (struct loop_exit): New function.
1725 (struct loop): single_exit_ field replaced by exits field.
1726 (LOOPS_HAVE_MARKED_SINGLE_EXITS): Replaced by LOOPS_HAVE_RECORDED_EXITS.
1727 (struct loops): Added exits hash.
1728 (mark_single_exit_loops, set_single_exit): Declaration removed.
1729 (release_recorded_exits, record_loop_exits, rescan_loop_exit): Declare.
1731 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1733 * doc/invoke.texi: Avoid use of @headitem.
1735 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1737 * cse.c (cse_insn): Move HAVE_CC0 code after declarations.
1739 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1741 * doc/install.texi: Fix m68k-*-* anchor and add m68k-*-* to the
1744 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1745 Richard Sandiford <richard@codesourcery.com>
1747 * doc/invoke.texi: Document -mno-bitfield, -mno-rtd and -mno-short.
1748 * config/m68k/m68k.opt: Resort options.
1749 (mbitfield, mrtd, mshort): Remove RejectNegative properties.
1751 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1752 Richard Sandiford <richard@codesourcery.com>
1754 * doc/invoke.texi: Document the macros that are defined by
1755 m68k's -mtune and -mhard-float options.
1756 * config/m68k/m68k-protos.h (m68k_cpp_cpu_ident) Declare.
1757 (m68k_cpp_cpu_family): Likewise.
1758 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Add a full set
1759 of __ucfv*__ macros. Define __mcffpu__ if generating code for
1760 ColdFire FPUs. Define __mcf_cpu_* and __mcf_family_* macros.
1761 * config/m68k/m68k.c (m68k_cpp_cpu_ident): New function.
1762 (m68k_cpp_cpu_family): Likewise.
1764 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1766 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Treat all mc68*
1767 macros besides mc68000 as tuning macros. Use a switch statement
1768 to set them and mcpu32.
1770 2007-01-12 Julian Brown <julian@codesourcery.com>
1772 * config/m68k/m68k.h: Use TARGET_68040 instead of TARGET_68040_ONLY.
1773 (TARGET_68040_ONLY): Rename to...
1774 (TARGET_68040): ...this.
1775 * config/m68k/m68k.c: Use TARGET_68040 instead of TARGET_68040_ONLY.
1776 * config/m68k/m68k.md: Likewise.
1778 2007-01-12 Julian Brown <julian@codesourcery.com>
1779 Nathan Sidwell <nathan@codesourcery.com>
1780 Richard Sandiford <richard@codesourcery.com>
1782 * config.gcc (m680[012]0-*-*, m68k*-*-*): Set m68k_cpu_ident to
1783 the -mcpu= argument associated with the --with-cpu setting.
1784 Define M68K_DEFAULT_TUNE to the default -mtune= option,
1785 if different from the one implied by the -mcpu setting.
1786 Accept --with-cpu=FOO if FOO is listed in m68k-devices.def,
1787 using mcpu=FOO as the default CPU option. Set target_cpu_default2.
1788 * doc/invoke.texi: Mention ColdFire in the introduction to the
1789 m68k options. Document the new -march, -mcpu, -mtune, -mdiv,
1790 -mno-div and -mhard-float options. Make -m68881 a synonym for
1791 -mhard-float. Document the previously-undocumented -m5206e,
1792 -m528x, -m5307 and -m5407 options. Tweak the existing option
1793 documentation for consistency.
1794 * doc/install.texi: Mention new --with-cpu arguments.
1795 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Only use the
1796 default CPU if neither -mcpu nor -march are specified.
1797 (ASM_CPU_SPEC): Pass down -mcpu and -march options.
1798 (TARGET_CPU_CPP_BUILTINS): Set __mcfisa*__ macros from
1799 TARGET_ISA*. Set the legacy __mcf*__ cpu macros in the same way,
1800 using m68k_tune to decide between families that implement the
1801 same ISA. Use m68k_tune to set __mcfv4e__.
1802 (FL_BITFIELD, FL_68881, FL_COLDFIRE, FL_CF_HWDIV, FL_CF_MAC)
1803 (FL_CF_EMAC, FL_CF_EMAC_B, FL_CF_USP, FL_CF_FPU, FL_ISA_68000)
1804 (FL_ISA_68010, FL_ISA_68020, FL_ISA_68040, FL_ISA_A, FL_ISA_B)
1805 (FL_ISA_C, FL_ISA_MMU): New macros.
1806 (MASK_COLDFIRE): Delete.
1807 (TARGET_68010, TARGET_68020, TARGET_68040_ONLY, TARGET_COLDFIRE)
1808 (TARGET_ISAB): Redefine in terms of m68k_cpu_flags.
1809 (TARGET_68881, TARGET_COLDFIRE_FPU): Redefine in terms of m68k_fpu.
1810 (TARGET_HARD_FLOAT): Do not define here.
1811 (TARGET_ISAAPLUS, TARGET_ISAC): New macros.
1812 (TUNE_68000): New macro.
1813 (TUNE_68000_10): Redefine in terms of TUNE_68000 and TUNE_68010.
1814 (TUNE_68010, TUNE_68030, TUNE_68040, TUNE_68060, TUNE_CPU32)
1815 (TUNE_CFV2): Redefine in terms of m68k_tune.
1816 (uarch_type, target_device, fpu_type): New enums.
1817 (m68k_cpu, m68k_tune, m68k_fpu, m68k_cpu_flags): Declare.
1818 * config/m68k/m68k.c (TARGET_DEFAULT): Remove MASK_68881.
1819 (FL_FOR_isa_00, FL_FOR_isa_10, FL_FOR_isa_20, FL_FOR_isa_40)
1820 (FL_FOR_isa_cpu32, FL_FOR_isa_a, FL_FOR_isa_aplus, FL_FOR_isa_b)
1821 (FL_FOR_isa_c): New macros.
1822 (m68k_isa): New enum.
1823 (m68k_target_selection): New structure.
1824 (all_devices, all_isas, all_microarchs): New tables.
1825 (m68k_cpu_entry, m68k_arch_entry, m68k_tune_entry, m68k_cpu)
1826 (m68k_tune, m68k_fpu, m68k_cpu_flags): New variables.
1827 (MASK_ALL_CPU_BITS): Delete.
1828 (m68k_find_selection): New function.
1829 (m68k_handle_option): Handle -mcpu=, -march= and -mtune=.
1830 Map the legacy target options to a combination of the new ones.
1831 (override_options): Set m68k_cpu, m68k_tune, m68k_fpu and
1832 m68k_cpu_flags. Handle M68K_DEFAULT_TUNE. Use m68k_cpu_flags
1833 to derive default MASK_BITFIELD, MASK_CF_HWDIV and MASK_HARD_FLOAT
1835 * config/m68k/m68k.opt (m5200, m5206e, m528x, m5307, m5407, mcfv4e)
1836 (m68010, m68020, m68020-40, m68020-60, m68030, m68040): Remove Mask
1838 (m68881, msoft-float): Change mask from 68881 to HARD_FLOAT.
1839 (march=, mcpu=, mdiv, mhard-float, mtune=): New options.
1840 * config/m68k/m68k-devices.def: New file.
1842 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1843 Nathan Sidwell <nathan@codesourcery.com>
1845 * config/m68k/m68k.h (ASM_CPU_SPEC, ASM_SPEC, EXTRA_SPECS)
1846 (SUBTARGET_EXTRA_SPECS): New macros.
1847 * config/m68k/linux.h (ASM_SPEC): Remove CPU flags;
1848 use %(asm_cpu_spec) instead.
1849 * config/m68k/m68k-none.h (ASM_SPEC): Likewise.
1850 * config/m68k/openbsd.h (ASM_SPEC): Likewise.
1851 * config/m68k/netbsd-elf.h (ASM_SPEC): Likewise.
1852 (EXTRA_SPECS): Rename to...
1853 (SUBTARGET_EXTRA_SPECS): ...this.
1855 2007-01-12 Nathan Sidwell <nathan@codesourcery.com>
1856 Richard Sandiford <richard@codesourcery.com>
1857 Julian Brown <julian@codesourcery.com>
1859 * config.gcc (m68k-*-aout*, m68k-*-coff*, m68020-*-elf*, m68k-*-elf*)
1860 (m68k-*-uclinux*, m68k-*-linux*, m68k-*-rtems*): Set default_m68k_cpu
1861 to the configuration's default CPU.
1862 (m68010-*-netbsdelf*, m68k*-*-netbsdelf*, m68k*-*-openbsd*): Likewise.
1863 Remove default masks.
1864 (m680[012]0-*-*): Set the default with_cpu to the first part of
1866 (m68k*-*-*): Set the default with_cpu to m$default_m68k_cpu.
1867 (m68k*-*-linux): Extend the --with-cpu handling to...
1868 (m680[012]0-*-*, m68k*-*-*): ...these configurations. Allow m68000
1869 and m68010. Don't set target_cpu_default2.
1870 * doc/install.texi: Document --with-cpu for m68k.
1871 * config/m68k/m68k.h (OPTION_DEFAULT_SPECS): Define.
1872 * config/m68k/m68k-none.h (TARGET_CPU_DEFAULT, M68K_CPU_m68k)
1873 (M68K_CPU_m68000, M68K_CPU_m68010, M68K_CPU_m68020, M68K_CPU_m68030)
1874 (M68K_CPU_m68040, M68K_CPU_m68302, M68K_CPU_m68332, TARGET_DEFAULT)
1875 (ASM_CPU_DEFAULT_SPEC, CC1_CPU_DEFAULT_SPEC): Delete.
1876 (ASM_SPEC): Remove use of %(asm_cpu_default).
1877 (EXTRA_SPECS, SUBTARGET_EXTRA_SPECS, MULTILIB_DEFAULTS): Delete.
1878 * config/m68k/linux.h (TARGET_DEFAULT): Delete.
1879 (CPP_SPEC): Merge definitions. Do not handle __HAVE_68881__ here.
1880 * config/m68k/netbsd-elf.h (TARGET_OS_CPP_BUILTINS): Define
1881 __HAVE_FPU__ if TARGET_HARD_FLOAT.
1882 (TARGET_DEFAULT): Delete.
1883 (EXTRA_SPECS): Delete cpp_cpu_default_spec, cpp_cpu_spec,
1884 cpp_fpu_spec, asm_default_spec and netbsd_cpp_spec.
1885 (CPP_CPU_SPEC): Delete.
1886 (TARGET_VERSION): Merge definitions, using TARGET_68010 to pick
1887 the appropriate string.
1888 (CPP_CPU_DEFAULT_SPEC, ASM_DEFAULT_SPEC, CPP_FPU_SPEC): Delete.
1889 (CPP_SPEC): Define to NETBSD_CPP_SPEC.
1890 (ASM_SPEC): Don't use %(asm_default_spec).
1891 * config/m68k/m68k.c (TARGET_DEFAULT_TARGET_FLAGS): Remove
1892 TARGET_DEFAULT and add MASK_68881.
1893 * config/m68k/m68k.md: Remove mention of TARGET_DEFAULT from comments.
1895 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1897 * config.gcc (m68010-*-netbsdelf*): Add MASK_68010.
1898 (m68k*-*-netbsdelf*, m68k*-*-openbsd*, m68k*-linux*): Add
1899 MASK_68010 alongside MASK_68020.
1900 * doc/invoke.texi: Document -m68010.
1901 * config/m68k/m68k.opt (m68010): New.
1902 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Define mc68010
1904 (TUNE_68010): New macro.
1905 * config/m68k/m68k-none.h (M68K_CPU_m68k, M68K_CPU_m68010)
1906 (M68K_CPU_m68020, M68K_CPU_m68030, M68K_CPU_m68040)
1907 (M68K_CPU_m68332): Add MASK_68010.
1908 * config/m68k/linux.h (TARGET_DEFAULT): Add MASK_68010 to
1909 fallback definition.
1910 * config/m68k/netbsd-elf.h (CPP_CPU_SPEC): Remove now-redundant
1912 * config/m68k/m68k.c (MASK_ALL_CPU_BITS): Add MASK_68010.
1913 (m68k_handle_option): Handle OPT_m68010. Add MASK_68010
1914 to all entries that use MASK_68020.
1915 (output_move_simode_const, output_move_himode, output_move_qimode)
1916 (output_move_stricthi, output_move_strictqi): Use TARGET_68010
1917 instead of TARGET_68020 to select clr behavior. Remove comment
1918 about there being no TARGET_68010.
1919 * config/m68k/m68k.md: Likewise throughout.
1921 2007-01-12 Julian Brown <julian@codesourcery.com>
1923 * config/m68k/m68k.h (TARGET_ISAB): New macro.
1924 * config/m68k/m68k.c: Use TARGET_ISAB rather than TARGET_CFV4.
1925 * config/m68k/m68k.md: Likewise.
1927 2007-01-12 Julian Brown <julian@codesourcery.com>
1929 * config/m68k/m68k.h (LEGITIMATE_INDEX_P, LEGITIMIZE_ADDRESS): Use
1930 TARGET_COLDFIRE_FPU instead of TARGET_CFV4E.
1932 2007-01-12 Julian Brown <julian@codesourcery.com>
1934 * config/m68k/m68k.h (TUNE_68040_60): New macro.
1935 * config/m68k/m68k.c (standard_68881_constant_p): Use it.
1936 * config/m68k/m68k.md: Likewise.
1938 2007-01-12 Julian Brown <julian@codesourcery.com>
1939 Richard Sandiford <richard@codesourcery.com>
1941 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Use TUNE_68030
1942 instead of TARGET_68030, TUNE_68040 instead of TARGET_68040,
1943 TUNE_68060 instead of TARGET_68060 and TUNE_CPU32 instead of
1945 (TARGET_CPU32): Rename to...
1946 (TUNE_CPU32): ...this.
1947 (TUNE_68000_10, TUNE_68030, TUNE_68040, TUNE_68060)
1948 (TUNE_CFV2): New macros.
1949 * config/m68k/netbsd-elf.h (LONG_DOUBLE_TYPE_SIZE): Simplify;
1950 remove conditions that are implied by TARGET_68020.
1951 * config/m68k/m68k.c (m68k_output_function_prologue): Use TUNE_68040
1952 instead of TARGET_68040 and TUNE_CPU32 instead of TARGET_CPU32.
1953 (m68k_output_function_epilogue): Likewise.
1954 (m68k_rtx_costs): Likewise. Use TUNE_68060 instead of TARGET_68060
1955 and TUNE_CFV2 instead of TARGET_5200. Use TUNE_68000_10 instead of
1956 "!TARGET_68020 && !TARGET_COLDFIRE" to choose between 68000 and
1957 non-68000 timings. Refactor multiplication and division costs.
1958 (output_addsi3): Use TUNE_68040 instead of TARGET_68040 and
1959 TUNE_CPU32 instead of TARGET_CPU32.
1960 (standard_68881_constant_p): Use TUNE_68040 instead of TARGET_68040
1961 and TUNE_68060 instead of TARGET_68060.
1962 * config/m68k/m68k.md: Use TUNE_68040 instead of TARGET_68040,
1963 TUNE_68060 instead of TARGET_68060, and TUNE_CPU32 instead of
1965 (movsi_const0): Use TUNE_68000_10 rather than "!TARGET_68020
1966 && !TARGET_COLDFIRE" to choose between moveq and clr.
1967 Likewise in the unnamed movsf pattern.
1968 (ashlsi_17_24, lshrsi_17_24): Guard with TUNE_68000_10 rather than
1969 "!TARGET_68020 && !TARGET_COLDFIRE". Likewise the unnamed
1972 2007-01-12 Richard Sandiford <richard@codesourcery.com>
1974 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Increase amount
1975 of tabbing before backslashes.
1977 2007-01-11 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1979 * pa-linux.h (ASM_OUTPUT_INTERNAL_LABEL): Undefine.
1980 * pa.h (ASM_OUTPUT_LABEL): Output colon when using GAS.
1981 (ASM_OUTPUT_INTERNAL_LABEL): Define.
1983 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1985 * tree-ssa-loop-ivopts.c (extract_cond_operands): Split from
1986 find_interesting_uses_cond.
1987 (find_interesting_uses_cond): Use extract_cond_operands.
1988 (rewrite_use_compare): Use extract_cond_operands and
1989 force_gimple_operand_bsi. Do not call update_stmt.
1990 (determine_use_iv_cost_condition): Use extract_cond_operands.
1991 Return cheaper of using original bound and changing the exit bound.
1993 2007-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1995 PR tree-optimization/29516
1996 * tree-ssa-address.c (tree_mem_ref_addr, add_to_parts,
1997 most_expensive_mult_to_index, addr_to_parts,
1998 create_mem_ref, maybe_fold_tmr): Make the type of
1999 fields of TARGET_MEM_REF sizetype.
2000 (move_fixed_address_to_symbol, move_pointer_to_base):
2002 * tree.def (TARGET_MEM_REF): Add comment on types of
2005 2007-01-11 Joseph Myers <joseph@codesourcery.com>
2007 * c-common.c (vector_types_convertible_p): Treat opaque types as
2008 always convertible if they have the same size, but not otherwise.
2010 2007-01-11 Steven Bosscher <steven@gcc.gnu.org>
2012 * ifcvt.c (struct noce_if_info): Add comments to the fields.
2013 Remove the b_unconditional field.
2014 (noce_try_sign_mask): Do not look at b_unconditional.
2015 (noce_process_if_block): Do not use merge_if_blocks. Update
2016 the CFG here. Do not set b_unconditional.
2017 (cond_move_process_if_block): Likewise.
2018 (find_cond_trap): Likewise.
2019 (check_cond_move_block): Require simple jump insns at the end
2022 2007-01-11 Jan Hubicka <jh@suse.cz>
2024 PR tree-optimization/1046
2025 * tree-tailcall.c (suitable_for_tail_call_opt_p): Use TREE_ADDRESSABLE
2026 when alias info is not ready.
2027 (pass_tail_recursion): Do not require aliasing.
2028 * tree-ssa-copyrename.c (pass_rename_ssa_cop): Likewise.
2029 * tree-ssa-ccp.c (pass_ccp, pass_fold_builtins): Likewise.
2030 * tree-ssa-copy.c (pass_copy_prop): Likewise.
2031 * tree-ssa-forwprop.c (pass_forwprop): Likewise.
2032 * tree-ssa-dce.c (pass_dce, pass_dce_loop, pass_cd_dce): Likewise.
2033 * passes.c (init_optimization_passes): Execute rename_ssa_copies,
2034 ccp, forwprop, copy_prop, merge_phi, copy_prop, dce and tail recursion
2036 * tree-ssa-operands.c (add_virtual_operand, get_indirect_ref_operand):
2037 When aliasing is not build, mark statement as volatile.
2039 2007-01-11 Tom Tromey <tromey@redhat.com>
2041 PR preprocessor/15185, PR preprocessor/20989:
2042 * doc/cppopts.texi <-MT>: Update description of algorithm for
2043 computing default target.
2044 <-M, -MD>: Reword "basename" text.
2046 2007-01-11 Roger Sayle <roger@eyesopen.com>
2048 * builtins.c (expand_builtin_pow, expand_builtin_powi,
2049 fold_builtin_cabs, fold_builtin_sqrt, fold_builtin_trunc,
2050 fold_builtin_floor, fold_builtin_ceil, fold_builtin_round,
2051 fold_builtin_int_int_roundingfn, fold_builtin_bitop,
2052 fold_builtin_bswap, real_constp, fold_builtin_pow,
2053 fold_builtin_powi, fold_builtin_signbit, fold_builtin_copysign,
2054 do_mpfr_arg1, do_mpfr_arg2, do_mpfr_arg3, do_mpfr_sincos): Replace
2055 uses of the macro TREE_CONSTANT_OVERFLOW with TREE_OVERFLOW.
2056 * convert.c (convert_to_pointer): Likewise.
2057 * expr.c (highest_pow2_factor, expand_expr_real_1): Likewise.
2058 * fold-const.c (force_fit_type, fold_negate_expr, int_const_binop,
2059 const_binop, fold_convert_const_int_from_int,
2060 fold_convert_const_int_from_real,
2061 fold_convert_const_real_from_real, sign_bit_p,
2062 optimize_minmax_comparison, extract_muldiv_1, fold_div_compare,
2063 fold_sign_changed_comparison, fold_unary, fold_comparison,
2064 fold_binary, multiple_of_p, tree_Expr_non_zero_p,
2065 fold_negate_const, fold_abs_const, fold_not_const): Likewise.
2066 * print-tree.c (print_node_brief, print_node): Likewise.
2067 * stor-layout.c (place_field, layout_type): Likewise.
2068 * tree-chrec.c (keep_cast): Likewise.
2069 * tree.c (build_vector, build_real, build_real_from_int_cst,
2070 build_complex): Likewise.
2072 2007-01-11 Roger Sayle <roger@eyesopen.com>
2074 * tree.h (TREE_CONSTANT_OVERFLOW): Obsolete. For the time being,
2075 treat TREE_CONSTANT_OVERFLOW as a synonym of TREE_OVERFLOW.
2077 2007-01-11 Paolo Bonzini <bonzini@gnu.org>
2079 * configure.ac (strict1_warn): Rename to strict_warn.
2080 (WERROR, --enable-werror, symlink hacks, stage1_cflags,
2081 cc_set_by_configure, quoted_cc_set_by_configure,
2082 stage_prefix_set_by_configure, quoted_stage_prefix_set_by_configure,
2083 all_boot_languages, all_stagestuff): Remove.
2084 (target_list): Remove bootstrap targets.
2085 * Makefile.in (quickstrap): Unconditionally make a synonym of all.
2086 (BOOT_LANGUAGES, STAGE1_CFLAGS, STAGE1_CHECKING,
2087 REMAKEFLAGS, FLAGS_TO_PASS, PREPEND_DOTDOT_TO_RELATIVE_PATHS,
2088 SUBDIR_FLAGS_TO_PASS, WERROR_FLAGS, STRICT2_WARN, LANG_STAGESTUFF,
2089 VOL_FILES, POSTSTAGE1_FLAGS_TO_PASS, STAGE2_FLAGS_TO_PASS,
2090 STAGEPROFILE_FLAGS_TO_PASS, STAGEFEEDBACK_FLAGS_TO_PASS, stage1_build,
2091 stage1_copy, stage2_build, stage2_copy, stageprofile_build,
2092 stageprofile_copy, stage3_build, stage3_copy, stagefeedback_build,
2093 stagefeedback_copy, stage4_build, clean_s1, clean_sw, bootstrap,
2094 bootstrap-lean, bootstrap2, bootstrap2-lean, bootstrap3,
2095 bootstrap3-lean, bootstrap4, bootstrap4-lean, unstage1, unstage2,
2096 unstage3, unstage4, unstageprofile, unstagefeedback, restage, restage2,
2097 restage3, restage4, restageprofile, restagefeedback, bubbleestrap,
2098 cleanstrap, unstrap, restrap, *compare, *compare3, *compare4,
2099 *compare-lean, *compare3-lean, *compare4-lean, stage1-start, stage1,
2100 stage2-start, stage2, stage3-start, stage3, stage4-start, stage4,
2101 stageprofile-start, stageprofile, stagefeedback-start, stagefeedback,
2102 risky-stage1, risky-stage2, risky-stage3, risky-stage4): Remove.
2103 (ORDINARY_FLAGS_TO_PASS): Rename to FLAGS_TO_PASS.
2104 (STAGECOPYSTUFF, STAGEMOVESTUFF): Consolidate into MOSTLYCLEANFILES.
2105 (mostlyclean): Adjust.
2106 (clean, distclean): Don't mention bootstrap stuff.
2107 * configure: Regenerate.
2108 * ada/config-lang.in, cp/config-lang.in, forttran/config-lang.in,
2109 java/config-lang.in, objc/config-lang.in, objcp/config-lang.in,
2110 treelang/config-lang.in (stagestuff): Remove.
2111 * doc/sourcebuild.texi (stage1, stage2, stage3, stage4,
2112 stageprofile, stagefeedback, stagestuff): Remove mention.
2114 2007-01-11 Nick Clifton <nickc@redhat.com>
2116 * config/mcore/predicates.md (mcore_general_movesrc_operand):
2118 (mcore_general_movdst_operand): Do not accept CONST_INTs.
2119 (mcore_arith_K_S_operand): Run the test for the S constraint not
2120 the test for the M constraint.
2121 (mcore_addsub_operand): Do not accept integer values that are
2122 larger than 32 bits.
2123 * config/mcore/mcore.md: Remove unused constraints from split.
2124 (andsi3): Use HOST_WIDE_INT instead of int to hold an INTVAL.
2126 (allocate_stack): Likewise.
2127 * config/mcore/mcore.c (mcore_print_operand): Restrict output of P
2128 operands to 32 bits.
2129 (mcore_const_costs): Use HOST_WIDE_INT instead of int to hold an
2131 (mcore_and_cost, mcore_modify_comparison, const_ok_for_mcore,
2132 mcore_const_ok_for_inline, mcore_const_trick_uses_not,
2133 try_constant_tricks, mcore_num_ones, mcore_num_zeros,
2134 mcore_output_bclri, mcore_output_andn, output_inline_const,
2135 mcore_output_move, mcore_output_movedouble): Likewise.
2136 (mcore_output_cmov): Use CONST_OK_FOR_M and CONST_OK_FOR_N.
2137 (output_inline_const): Likewise.
2138 (output_inline_const): Fix format strings used in sprintf
2140 * config/mcore/mcore-protos.h: Update prototypes for changed
2141 functions in mcore.c.
2142 * config/mcore/mcore.h (CONST_OK_FOR_I): Cast values to
2143 HOST_WIDE_INT and not int.
2144 (CONST_OK_FOR_J, CONST_OK_FOR_K, CONST_OK_FOR_L, CONST_OK_FOR_M,
2145 CONST_OK_FOR_N): Likewise.
2146 (LEGITIMATE_CONSTANT_P): Also check CONSTANT_P.
2147 (GO_IF_LEGITIMATE_INDEX): Use HOST_WIDE_INT instead of int to hold
2150 2007-01-10 Jan Hubicka <jh@suse.cz>
2152 * tree-vrp.c (remove_range_assertions): Release defs.
2153 * tree-ssa-loop-ivopts.c (rmeove_statement): Likewise.
2154 * tree-ssa-dom.c (remove_stmt_or_phi): Likewise.
2156 2007-01-10 Paul Brook <paul@codesourcery.com>
2158 * config/arm/arm.c (arm_rtx_costs_1): Handle mutiply-subtract.
2159 * config/arm/arm.md (mulsi3subsi): New insn.
2161 2007-01-10 Zdenek Dvorak <dvorakz@suse.cz>
2163 * tree-ssa-loop-manip.c (tree_unroll_loop): Make it a wrapper over ...
2164 (tree_transform_and_unroll_loop): New.
2165 * tree-flow.h (transform_callback, tree_transform_and_unroll_loop):
2168 2007-01-10 Robert Kennedy <jimbob@google.com>
2170 * fold-const.c (fold_comparison): Fold comparisons like (x *
2171 1000 < 0) to (x < 0).
2173 2007-01-10 Ian Lance Taylor <iant@google.com>
2175 * tree-pretty-print.c (dump_generic_node): Print parentheses when
2176 operands have the same priority.
2178 2007-01-10 Tom Tromey <tromey@redhat.com>
2180 * fold-const.c (fold_truthop): Don't check can_use_bit_fields_p.
2181 (fold_binary): Likewise.
2182 * langhooks.c (lhd_can_use_bit_fields_p): Removed.
2183 * langhooks-def.h (lhd_can_use_bit_fields_p): Removed.
2184 (LANG_HOOKS_CAN_USE_BIT_FIELDS_P): Removed.
2185 (LANG_HOOKS_INITIALIZER): Remove LANG_HOOKS_CAN_USE_BIT_FIELDS_P.
2186 * langhooks.h (struct lang_hooks): Removed field
2187 'can_use_bit_fields_p'.
2189 2007-01-10 Ralf Corsépius <ralf.corsepius@rtems.org>
2191 * config/bfin/t-bfin, config/bfin/t-bfin-elf: Remove GCC_CFLAGS.
2193 2007-01-10 Razya Ladelsky <razya@il.ibm.com>
2195 * function.c (get_last_funcdef_no): New function.
2196 * function.h (get_last_funcdef_no): Declare.
2197 * tree-inline.c (initialize_cfun): Add initialization.
2198 (tree_function_versioning): Cleanup.
2200 2007-01-10 Jan Hubicka <jh@suse.cz>
2202 * tree-inline.c (setup_one_parameter): Do not propagate into abnormal
2205 2007-01-10 Sa Liu <saliu@de.ibm.com>
2206 Ben Elliston <bje@au.ibm.com>
2208 * spu.h (STACK_SAVE_AREA): Use VOIDmode for SAVE_FUNCTION, SImode
2209 for SAVE_NONLOCAL and Pmode for any other save level.
2210 * spu-protos.h (spu_restore_stack_block): Declare.
2211 * spu.md (save_stack_block): Remove.
2212 (restore_stack_block): Call spu_restore_stack_block.
2213 * spu.c (spu_restore_stack_block): New function.
2214 (spu_expand_epilogue): Remove old comment.
2216 2007-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2218 PR tree-optimization/30322
2219 * tree-ssa-loop-ivopts.c (fold_affine_expr, iv_value): Removed.
2220 (cand_value_at): Return the value as aff_tree.
2221 (may_eliminate_iv): Convert the bound from aff_tree to tree.
2222 * tree-affine.c (aff_combination_add_cst, aff_combination_add_product,
2223 aff_combination_mult): New functions.
2224 (aff_combination_add): Use aff_combination_add_cst.
2225 (aff_combination_convert): Allow conversions to a wider type.
2226 (tree_to_aff_combination): Handle BIT_NOT_EXPR.
2227 * tree-affine.h (aff_combination_mult): Declare.
2229 2007-01-09 Carlos O'Donell <carlos@codesourcery.com>
2231 * doc/tm.texi: Update documentation to reflect reality of exec
2232 and start file search behaviours. Update copyright year.
2233 * doc/invoke.texi: Explain how GCC_EXEC_PREFIX is used to find
2234 header file directories.
2236 2007-01-09 Uros Bizjak <ubizjak@gmail.com>
2238 * config/i386/i386.md (*sinxf2): Rename to *sinxf2_i387.
2239 (*cosxf2): Rename to cosxf2_i387.
2240 (*sindf2, *sinsf2): Extend operand 1 to XFmode. Macroize patterns
2241 using X87MODEF12 mode macro. Rename patterns to
2242 *sin_extend<mode>xf2_i387. Use SSE_FLOAT_MODE_P to disable patterns
2244 (*cosdf2, *cossf2): Ditto.
2245 (sincosdf3, sincossf3): Ditto. Rewrite corresponding splitters
2246 to match extended input operands.
2247 (sincos<mode>3): New expander.
2248 (*sinextendsfdf2, *cosextendsfdf2, *sincosextendsfdf3): Remove
2249 insn patterns and corresponding splitters.
2251 2007-01-09 Kaz Kojima <kkojima@gcc.gnu.org>
2253 * config/sh/t-linux (TARGET_LIBGCC2_CFLAGS): Delete.
2254 (SHLIB_MAPFILES, SHLIB_LINK, SHLIB_INSTALL): Likewise.
2256 2007-01-09 Nicolas Pitre <nico@cam.org>
2259 * arm/ieee754-df.S (Lad_s): Also test the low word of X for zero.
2261 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2263 * target.h (struct gcc_target): New field library_rtti_comdat.
2264 * target-def.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): New.
2265 (TARGET_CXX): Add TARGET_CXX_LIBRARY_RTTI_COMDAT.
2266 * doc/tm.texi (C++ ABI): Document TARGET_CXX_LIBRARY_RTTI_COMDAT.
2267 * config/darwin.h (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
2269 2007-01-08 Geoffrey Keating <geoffk@apple.com>
2271 * doc/invoke.texi (Optimize Options): Correct description of -O0.
2273 2007-01-08 Richard Guenther <rguenther@suse.de>
2275 * tree.h (force_fit_type_double): Export.
2276 (force_fit_type): Remove.
2277 * fold-const.c (force_fit_type_double): New function.
2278 (force_fit_type): Remove.
2279 (int_const_binop): Use it.
2280 (fold_convert_const_int_from_int): Likewise.
2281 (fold_convert_const_int_from_real): Likewise.
2282 (fold_div_compare): Likewise.
2283 (fold_sign_changed_comparison): Likewise.
2284 (fold_unary): Likewise.
2285 (fold_negate_const): Likewise.
2286 (fold_abs_const): Likewise.
2287 (fold_not_const): Likewise.
2288 * c-common.c (shorten_compare): Use force_fit_type_double.
2289 * convert.c (convert_to_pointer): Likewise.
2291 2007-01-08 Richard Guenther <rguenther@suse.de>
2293 * tree.h (build_int_cst_wide_type): Export.
2294 * tree.c (build_int_cst_wide_type): New function.
2295 (build_int_cst_wide): Fix comment.
2296 * builtins.c (fold_builtin_object_size): Use build_int_cst
2297 to build -1 or 0 of the correct type. Use fit_double_type
2298 to check for overflow.
2299 * fold-const.c (optimize_bit_field_compare): Use build_int_cst_type
2301 (decode_field_reference): Likewise.
2302 (all_ones_mask_p): Likewise.
2303 (native_interpret_int): Use build_int_cst_wide_type.
2304 (fold_binary): Use build_int_cst_type to build an all-ones
2306 * stor-layout.c (set_sizetype): Use build_int_cst_wide_type.
2308 2007-01-08 Daniel Jacobowitz <dan@codesourcery.com>
2310 * config/pa/t-pa64 (libgcc_stub.a): Use $(T).
2312 2007-01-09 Ben Elliston <bje@au.ibm.com>
2314 * genautomata.c (STATS_OPTION): New option.
2315 (stats_flag): New flag.
2316 (gen_automata_option): Handle it.
2317 (initiate_automaton_gen): Ditto.
2318 (write_automata): Output statistics only if stats_flag is
2319 set. Likewise, output time statistics only if time_flag is set.
2320 * doc/md.texi (Processor pipeline description): Document new flag.
2322 2007-01-08 Richard Guenther <rguenther@suse.de>
2324 * builtins.c (fold_builtin_int_roundingfn): Use fit_double_type.
2325 * tree.c (build_int_cst_type): Likewise.
2326 (size_in_bytes): Don't call force_fit_type on the result.
2327 (int_fits_type_p): Use fit_double_type.
2328 * fold-const.c (fit_double_type): New function.
2329 (force_fit_type): Use it.
2330 * tree.h (fit_double_type): Export.
2332 2007-01-08 Jan Hubicka <jh@suse.cz>
2334 * tree-vectorizer.c (gate_increase_alignment): Fix return type.
2335 * ipa.c (function_and_variable_visibility): Fix return type.
2337 2007-01-08 Richard Guenther <rguenther@suse.de>
2339 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Use type
2340 of offset to build the index.
2341 * tree-pretty-print.c (dump_generic_node): Don't build negated
2342 const just for printing.
2343 * c-pretty-print.c (pp_c_integer_constant): Likewise.
2344 * builtins.c (fold_builtin_int_roundingfn): Check if result
2345 fits the type by using force_fit_type and comparing the result.
2346 * predict.c (predict_loops): Use compare_tree_int for comparison.
2347 * tree.c (build_int_cst): Fall back to integer_type_node for
2349 (build_int_cst_wide): Assert type is non-null.
2351 2007-01-08 Roberto Costa <roberto.costa@st.com>
2353 * tree-vrp.c (extract_range_from_cond_expr): New.
2354 (extract_range_from_expr): Handle COND_EXPR nodes used as expressions.
2355 * tree-ssa-ccp.c (get_maxval_strlen): Handle COND_EXPR nodes used
2357 (fold_stmt): Bug fix, avoid infinite recursion when folding COND_EXPRs.
2358 * tree-ssa-forwprop.c (simplify_cond, forward_propagate_into_cond,
2359 tree_ssa_forward_propagate_single_use_vars): Handle COND_EXPR nodes
2360 used as expressions.
2361 * tree-object-size.c (cond_expr_object_size): New.
2362 (collect_object_sizes_for): Handle COND_EXPR nodes used as expressions.
2364 2007-01-08 Jan Hubicka <jh@suse.cz>
2366 * tree-ssa-forwprop.c (forward_propagate_into_cond,
2367 tree_ssa_forward_propagate_single_use_va): Release defs of propagated
2370 2007-01-08 Richard Guenther <rguenther@suse.de>
2372 PR tree-optimization/23603
2373 * tree-vrp.c (set_value_range_to_truthvalue): New function.
2374 (extract_range_from_binary): Fall back to truthvalue instead of
2375 varying for TRUTH_*_EXPR.
2376 (extract_range_from_comparison): Fall back to truthvalue instead of
2378 (vrp_visit_phi_node): Don't adjust new range bounds to +INF/-INF
2379 if all visited PHI values were constant.
2381 2007-01-08 Jan Hubicka <jh@suse.cz>
2383 * cgraphunit.c (cgraph_process_new_functions): Reset reachable flag.
2384 (cgraph_analyze_function): break out from ...
2385 (cgraph_finalize_compilation_unit): ... here.
2386 (cgraph_expand_function): Remove forgoten commented out line.
2387 (cgraph_optimize): Analyze functions.
2389 2007-01-08 Jan Hubicka <jh@suse.cz>
2391 * tree-pas.h (TODO_remove_function): New flag.
2392 (TODO_update*): Renumber.
2393 (pass_ipa_increase_alignment,
2394 pass_ipa_function_and_variable_visibility): New passes.
2395 * cgraphunit.c (cgraph_increase_alignment): Move to tree-vectorizer.c
2396 (cgraph_function_and_variable_visibility): Move to ipa.c
2397 (cgraph_optimize): Don't call cgraph_function_and_variable_visibility,
2398 cgraph_increase_alignment.
2399 * ipa-inline.c (cgraph_decide_inlining): Don't push timevar.
2400 (cgraph_decide_inlining_incrementally): Push TV_INTEGRATION before
2401 calling tree-inline.
2402 (cgraph_early_inlining): Do not call cgraph_remove_unreachable_nodes.
2403 (pass_ipa_inline, pass_early_ipa_inlining): Set TODO_remove_functions
2404 * tree-vectorizer.c (increase_alignment): Move here from cgraphunit.c
2405 (gate_increase_alignment): New function.
2406 (pass_ipa_increase_alignment): New pass.
2407 * ipa.c: Inline tree-pass.h and timevar.h
2408 (function_and_variable_visibility): Move here from cgraphunit.c
2409 * tree-optimize.c (pass_early_local_passes): Add TODO_remove_functions.
2410 * passes.c (init_optimization_passes): Add the two new passes.
2411 (execute_todo): Handle cgraph_remove_functions.
2413 2007-01-08 Nick Clifton <nickc@redhat.com>
2415 * config/frv/predicates.md (reg_or_0_operand): Accept
2418 2007-01-08 Ralf Corsépius <ralf.corsepius@rtems.org>
2420 * config/bfin/rtems.h, config/bfin/t-rtems: New.
2421 * config.gcc: Add bfin*-rtems*.
2423 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2425 * c.opt: Add -flax-vector-conversions.
2426 * c-typeck.c (convert_for_assignment): Pass flag to
2427 vector_types_convertible_p to allow emission of note.
2428 (digest_init): Likewise.
2429 * c-opts.c: Handle -flax-vector-conversions.
2430 * c-common.c (flag_lax_vector_conversions): New.
2431 (vector_types_convertible_p): Unless -flax-vector conversions
2432 has been passed, disallow conversions between vectors with
2433 differing numbers of subparts and/or element types. If such
2434 a conversion is disallowed, possibly emit a note on the first
2435 occasion only to inform the user of -flax-vector-conversions.
2436 The new last argument specifies this.
2437 * c-common.h (flag_lax_vector_conversions): New.
2438 (vector_types_convertible_p): Add extra argument.
2439 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use
2440 char_type_node for V*QI type vectors.
2441 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
2442 Update to satisfy new typechecking rules.
2443 * config/rs6000/altivec.h (vec_cmple): Use vec_cmpge, for both
2445 * doc/invoke.texi (C Dialect Options): Document
2446 -flax-vector-conversions.
2448 2007-01-08 Mark Shinwell <shinwell@codesourcery.com>
2450 PR tree-optimization/29877
2451 * tree-ssa-ter.c (is_replaceable_p): Deem assignments with
2452 a register variable on the RHS to not be replaceable.
2454 2007-01-08 Chen Liqin <liqin@sunnorth.com.cn>
2455 * config/score/t-score-elf (MULTILIB_OPTIONS): Change.
2456 * config/score/predicates.md (const_uimm5, sr0_operand, const_simm12,
2457 const_simm15, const_pow2, const_npow2): Added.
2458 * config/score/misc.md (insv, extv, extzv, movmemsi,
2459 move_lbu_a/b, mov_lhu_a/b etc): Added and fix some bug.
2460 * config/score/score.c (score_address_cost, score_select_cc_mode):
2462 Change CONST_OK_FOR_LETTER_P/EXTRA_CONSTRAINT define.
2463 Update score_rtx_costs for MACRO TARGET_RTX_COSTS.
2464 Update score_print_operand.
2465 * config/score/score.h (DATA_ALIGNMENT, SELECT_CC_MODE): Added.
2466 Adjust register allocate order and update some macro define.
2467 * config/score/score-mdaux.c (mdx_unaligned_load, mdx_unsigned_store,
2468 mdx_block_move_straight, mdx_block_move_loop_head,
2469 mdx_block_move_loop_body, mdx_block_move_loop_foot, mdx_block_move_loop,
2470 mdx_block_move): Added.
2471 (mdx_movsicc, mdp_select_add_imm, mdp_select, mds_zero_extract_andi,
2472 mdp_limm): Updated and fix some bug and typo.
2473 * config/score/score.md (movqi/hi/si, add/sub/zero/ext): Updated.
2474 (movsf, movdf, doloop_end): Added.
2476 2007-01-08 Kazu Hirata <kazu@codesourcery.com>
2478 * config/arm/arm.c, config/arm/arm.h, config/arm/arm.md,
2479 config/arm/thumb2.md: Fix comment typos.
2480 * doc/extend.texi: Fix a typo.
2482 2007-01-07 Eric Christopher <echristo@apple.com>
2484 * configure.ac: Check for __stack_chk_fail for darwin.
2485 * configure: Regenerate.
2487 2007-01-07 Richard Guenther <rguenther@suse.de>
2489 * tree-vrp.c (extract_range_from_assert): CSE calls to
2490 compare_values where possible.
2491 (extract_range_from_unary_expr): Likewise.
2493 2007-01-07 Anatoly Sokolov <aesok@post.ru>
2495 * config/avr/avr-protos.h (call_insn_operand): Delete prototype.
2496 * config/avr/avr.c (call_insn_operand): Delete function.
2497 * config/avr/avr.md (*pushqi, *pushhi, *pushsi, *pushsf): Use REG_SP
2498 instead of register number. Use predicates.
2499 * config/avr/predicates.md (const0_operand, reg_or_0_operand,
2500 call_insn_operand): Add.
2502 2007-01-06 Jan Hubicka <jh@suse.cz>
2504 * tree-pass.h (pass_build_cgraph_edges): Declare.
2505 * cgraphunit.c (record_refernece): Move to cgraphbuild.c
2506 (visited_nodes): Remove.
2507 (cgraph_create_edges): Move to cgraphbuild.c; rename to
2508 build_cgrpah_edges; make visited_nodes local.
2509 (cgraph_process_new_functions): DO not call initialize_inline_failed.
2510 (record_references_in_initializer): Move to cgraphbuild.c
2511 (initialize_inline_failed, rebuild_cgraph_edges,
2512 pass_rebuild_cgraph_edges): Move to cgraphbuild.c.
2513 (verify_cgraph_node): Make visited_nodes local.
2514 (cgraph_analyze_function): Do not call cgraph_create_edges and
2515 initialize_inline_failed.
2516 (cgraph_expand_function): Do not call cgraph_lower_function;
2517 assert that function is already lowered.
2518 * Makefile.in (cgraphbuild.o): New.
2519 * passes.c (init_optimization_passes): Add pass_build_cgraph_edges
2520 at the end of lowering passes.
2522 2007-01-06 Steven Bosscher <steven@gcc.gnu.org>
2524 * ifcvt.c (cond_move_convert_if_block): New function, code
2525 factored out from...
2526 (cond_move_process_if_block): ...here. Call the new function
2527 on the THEN and ELSE blocks.
2528 (merge_if_block): Do not copy global_live_at_end, merge_blocks
2529 already takes care of this.
2531 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2534 * tree.h (TREE_OVERFLOW_P): New.
2535 * c-typeck.c (parser_build_unary_op): Warn only if result
2536 overflowed and operands did not.
2537 (parser_build_binary_op): Likewise.
2538 (convert_for_assignment): Remove redundant overflow_warning.
2539 * c-common.c (overflow_warning): Don't check or set TREE_OVERFLOW.
2541 2007-01-05 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2543 * c-typeck.c (store_init_value): Split over two lines to follow
2544 the GNU coding style.
2546 2007-01-05 Benjamin Kosnik <bkoz@redhat.com>
2548 * c-cppbuiltin.c (c_cpp_builtins): __GXX_EXPERIMENTAL_CPP0X__ to
2549 __GXX_EXPERIMENTAL_CXX0X__.
2550 * doc/cpp.texi: Same.
2552 2007-01-05 Richard Guenther <rguenther@suse.de>
2555 * tree.c (get_narrower): Do not construct COMPONENT_REFs
2556 with mismatched types. Instead explicitly build a
2557 conversion NOP_EXPR.
2559 2007-01-05 Ian Lance Taylor <iant@google.com>
2561 * c-common.c (decl_with_nonnull_addr_p): New function.
2562 (c_common_truthvalue_conversion): Call it.
2563 * c-typeck.c (build_binary_op): Likewise.
2564 * c-common.h (decl_with_nonnull_addr_p): Declare.
2566 2007-01-05 Jakub Jelinek <jakub@redhat.com>
2569 * libgcc2.c (__divdc3): Compare c and d against 0.0 instead of
2572 2007-01-05 Joel Brobecker <brobecker@adacore.com>
2574 * doc/install.texi (Final install): Document the fact that
2575 the GNAT runtime should not be stripped.
2577 2007-01-04 Jan Hubicka <jh@suse.cz>
2579 * tree-inline.c (fold_marked_statements): Update operand caches
2580 and EH after folding
2582 2007-01-04 Ian Lance Taylor <iant@google.com>
2584 * c-common.c (check_function_nonnull): Whitespace fix.
2586 2007-01-04 Jan Hubicka <jh@suse.cz>
2588 * tree-optimize.c (execute_fixup_cfg): Correct previously mistakely
2589 comitted older version of patch.
2590 (pass_fixup_cfg): Add TODOs to verify flow and statements, dump
2591 function, celanup cfg and collect garbage.
2593 2007-01-04 Mike Stump <mrs@apple.com>
2595 * Makefile.in (mostlyclean): Don't remove libgcc anymore.
2598 2007-01-04 Eric Christopher <echristo@apple.com>
2600 * libgcc2.c (__bswapsi2): Use SItype.
2601 (__bswapdi2): Use DItype.
2602 * libgcc2.h: Update for above.
2604 2007-01-04 Paul Brook <paul@codesourcery.com>
2606 * config/arm/arm.md (arm_mulsi3, thumb_mulsi3, mulsi3_compare0,
2607 mulsi_compare0_scratch, mulsi3addsi, mulsi3addsi_compare0,
2608 mulsi3addsi_compare0_scratch, mulsidi3adddi, mulsidi3,
2609 umulsidi3, umulsidi3adddi, smulsi3_highpart,
2610 umulsi3_highpart): Make conditional on !arm_arch6.
2611 (arm_mulsi3_v6, thumb_mulsi3_v6, mulsi3_compare0_v6,
2612 mulsi_compare0_scratch_v6, mulsi3addsi_v6, mulsi3addsi_compare0_v6,
2613 mulsi3addsi_compare0_scratch_v6, mulsidi3adddi_v6, mulsidi3_v6,
2614 umulsidi3_v6, umulsidi3adddi_v6, smulsi3_highpart_v6,
2615 umulsi3_highpart_v6): New insns.
2617 2007-01-04 Roger Sayle <roger@eyesopen.com>
2619 * fold-const.c (fold_convert): When casting an expression to void,
2620 fold_ignored_result may discover a GIMPLE_MODIFY_STMT which doesn't
2621 have a type. Instead of attempting to build a NOP_EXPR, return
2622 these "special" trees directly.
2624 2007-01-04 Joseph Myers <joseph@codesourcery.com>
2626 * config/rs6000/rs6000.c (rs6000_rtx_costs): Make adjustment for
2627 MULT inside MINUS as either argument. Use rs6000_cost->dmul -
2628 rs6000_cost->fp not 0 as adjustment for outer NEG.
2630 2007-01-04 Jan Hubicka <jh@suse.cz>
2632 * cgraph.c (cgraph_release_function_body): New function.
2633 (cgraph_remove_node): Use it.
2634 * cgraph.h (cgraph_release_function_body): Declare.
2635 * cgraphunit.c (cgraph_expand_function): Use it.
2636 * ipa.c (cgraph_remove_unreahchable_nodes): Use it.
2637 * tree-ssa.c (delete_tree_ssa): Allow to be called before aliasing
2638 is initialized and while compilation of other function is running.
2639 * tree-optimize.c (execute_free_cfg_annotations): Move code to clear
2640 statement CFG annotations from here to ...
2641 * tree-cfg.c (delete_tree_cfg_annotations): ... here.
2643 2007-01-04 Zdenek Dvorak <dvorakz@suse.cz>
2645 * cfgloop.h (enum li_flags): Make the constants powers of two.
2647 2007-01-04 Jan Hubicka <jh@suse.cz>
2649 * tree-inline.c (copy_bb): Insert new statements to statements_to_fold
2651 (fold_marked_statements): New function.
2652 (optimize_inline_calls, tree_function_versioning): Fold new statements.
2653 * tree-inline.h (copy_body_data): Add statements_to_fold.
2655 2007-01-03 Daniel Jacobowitz <dan@codesourcery.com>
2657 * config.gcc: Mention libgcc/config.host.
2658 * Makefile.in: Update comments mentioning libgcc.
2659 (LIBGCC, INSTALL_LIBGCC, GCC_PARTS, mklibgcc): Delete.
2660 (all.cross, start.encap, rest.encap, rest.cross): Update
2661 dependencies for libgcc move.
2662 (libgcc.mk, LIBGCC_DEPS, libgcov.a, libgcc.a, stmp-multilib)
2663 (clean-target, clean-target-libgcc): Delete.
2664 (srcdirify, GCC_EXTRA_PARTS): New macros.
2665 (libgcc-support, libgcc.mvars): New rules.
2666 (distclean): Remove mention of mklibgcc.
2667 (install): Don't reference INSTALL_LIBGCC.
2668 (install-common): Don't reference EXTRA_PARTS.
2669 (install-libgcc, install-multilib): Delete rules.
2670 * mklibgcc.in: Delete file.
2671 * doc/configfiles.texi: Don't mention mklibgcc.
2673 * config/i386/t-darwin (SHLIB_VERPFX): Delete (moved to libgcc).
2674 * config/i386/t-darwin64 (SHLIB_VERPFX): Likewise.
2675 * config/rs6000/t-darwin (SHLIB_VERPFX): Likewise.
2676 * config/rs6000/t-ppccomm (TARGET_LIBGCC2_CFLAGS, SHLIB_MAPFILES)
2677 (mklibgcc, ldblspecs): Likewise.
2679 * config/i386/t-nwld (libgcc.def, libc.def, libpcre.def)
2680 (posixpre.def): Use $(T).
2681 (SHLIB_EXT, SHLIB_NAME, SHLIB_SLIBDIR_QUAL, SHLIB_DEF, SHLIB_MAP)
2682 (SHLIB_SRC, SHLIB_INSTALL): Delete.
2683 (SHLIB_LINK): Make dummy.
2684 * config/t-slibgcc-darwin: Delete contents except for dummy SHLIB_LINK.
2686 * config/frv/t-linux (EXTRA_MULTILIB_PARTS): Clear.
2688 * config/alpha/t-crtfm: Use $(T) in rules for EXTRA_PARTS.
2689 * config/alpha/t-vms, config/alpha/t-vms64, config/fr30/t-fr30,
2690 config/i386/t-rtems-i386, config/ia64/t-ia64, config/rs6000/t-beos,
2691 config/rs6000/t-newas, config/sparc/t-elf: Likewise.
2693 * configure.ac (all_outputs): Remove mklibgcc.
2694 * configure: Regenerated.
2696 2007-01-03 Josh Conner <jconner@apple.com>
2699 * calls.c (compute_argument_addresses): Set stack and stack_slot
2700 for partial args, too.
2701 (store_one_arg): Use locate.size.constant for the size when
2702 generating a save_area.
2704 2007-01-03 Robert Kennedy <jimbob@google.com>
2706 * tree-cfg.c (tree_merge_blocks): Release SSA_NAME phi results
2707 whose definitions are deleted due to basic block merging.
2709 2007-01-03 Paul Brook <paul@codesourcery.com>
2712 * config/arm/arm.c (output_return_instruction): Pop PC in interrupt
2714 (use_return_insn): Return 0 for Thumb interrupt functions.
2715 (print_multi_reg): Add rfe argument for IRQ returns.
2716 (arm_output_epilogue): Pop interrupt return address directly into PC.
2717 (arm_expand_prologue): Only adjust IRQ return address in Arm mode.
2719 2007-01-03 Paul Brook <paul@codesourcery.com>
2721 Merge from sourcerygxx-4_1.
2722 * config/arm/thumb2.md: New file.
2723 * config/arm/elf.h (JUMP_TABLES_IN_TEXT_SECTION): Return True for
2725 * config/arm/coff.h (JUMP_TABLES_IN_TEXT_SECTION): Ditto.
2726 * config/arm/aout.h (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2727 (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump tables.
2728 * config/arm/aof.h (ASM_OUTPUT_ADDR_DIFF_ELT): Output Thumb-2 jump
2730 (ASM_OUTPUT_ADDR_VEC_ELT): Add !Thumb-2 assertion.
2731 * config/arm/ieee754-df.S: Use macros for Thumb-2/Unified asm
2733 * config/arm/ieee754-sf.S: Ditto.
2734 * config/arm/arm.c (thumb_base_register_rtx_p): Rename...
2735 (thumb1_base_register_rtx_p): ... to this.
2736 (thumb_index_register_rtx_p): Rename...
2737 (thumb1_index_register_rtx_p): ... to this.
2738 (thumb_output_function_prologue): Rename...
2739 (thumb1_output_function_prologue): ... to this.
2740 (thumb_legitimate_address_p): Rename...
2741 (thumb1_legitimate_address_p): ... to this.
2742 (thumb_rtx_costs): Rename...
2743 (thumb1_rtx_costs): ... to this.
2744 (thumb_compute_save_reg_mask): Rename...
2745 (thumb1_compute_save_reg_mask): ... to this.
2746 (thumb_final_prescan_insn): Rename...
2747 (thumb1_final_prescan_insn): ... to this.
2748 (thumb_expand_epilogue): Rename...
2749 (thumb1_expand_epilogue): ... to this.
2750 (arm_unwind_emit_stm): Rename...
2751 (arm_unwind_emit_sequence): ... to this.
2752 (thumb2_legitimate_index_p, thumb2_legitimate_address_p,
2753 thumb1_compute_save_reg_mask, arm_dwarf_handle_frame_unspec,
2754 thumb2_index_mul_operand, output_move_vfp, arm_shift_nmem,
2755 arm_save_coproc_regs, thumb_set_frame_pointer, arm_print_condition,
2756 thumb2_final_prescan_insn, thumb2_asm_output_opcode, arm_output_shift,
2757 thumb2_output_casesi): New functions.
2758 (TARGET_DWARF_HANDLE_FRAME_UNSPEC): Define.
2759 (FL_THUMB2, FL_NOTM, FL_DIV, FL_FOR_ARCH6T2, FL_FOR_ARCH7,
2760 FL_FOR_ARCH7A, FL_FOR_ARCH7R, FL_FOR_ARCH7M, ARM_LSL_NAME,
2761 THUMB2_WORK_REGS): Define.
2762 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv, arm_condexec_count,
2763 arm_condexec_mask, arm_condexec_masklen)): New variables.
2764 (all_architectures): Add armv6t2, armv7, armv7a, armv7r and armv7m.
2765 (arm_override_options): Check new CPU capabilities.
2766 Set new architecture flag variables.
2767 (arm_isr_value): Handle v7m interrupt functions.
2768 (user_return_insn): Return 0 for v7m interrupt functions. Handle
2770 (const_ok_for_arm): Handle Thumb-2 constants.
2771 (arm_gen_constant): Ditto. Use movw when available.
2772 (arm_function_ok_for_sibcall): Return false for v7m interrupt
2774 (legitimize_pic_address, arm_call_tls_get_addr): Handle Thumb-2.
2775 (thumb_find_work_register, arm_load_pic_register,
2776 legitimize_tls_address, arm_address_cost, load_multiple_sequence,
2777 emit_ldm_seq, emit_stm_seq, arm_select_cc_mode, get_jump_table_size,
2778 print_multi_reg, output_mov_long_double_fpa_from_arm,
2779 output_mov_long_double_arm_from_fpa, output_mov_double_fpa_from_arm,
2780 output_mov_double_fpa_from_arm, output_move_double,
2781 arm_compute_save_reg_mask, arm_compute_save_reg0_reg12_mask,
2782 output_return_instruction, arm_output_function_prologue,
2783 arm_output_epilogue, arm_get_frame_offsets, arm_regno_class,
2784 arm_output_mi_thunk, thumb_set_return_address): Ditto.
2785 (arm_expand_prologue): Handle Thumb-2. Use arm_save_coproc_regs.
2786 (arm_coproc_mem_operand): Allow POST_INC/PRE_DEC.
2787 (arithmetic_instr, shift_op): Use arm_shift_nmem.
2788 (arm_print_operand): Use arm_print_condition. Handle '(', ')', '.',
2790 (arm_final_prescan_insn): Use extract_constrain_insn_cached.
2791 (thumb_expand_prologue): Use thumb_set_frame_pointer.
2792 (arm_file_start): Output directive for unified syntax.
2793 (arm_unwind_emit_set): Handle stack alignment instruction.
2794 * config/arm/lib1funcs.asm: Remove default for __ARM_ARCH__.
2795 Add v6t2, v7, v7a, v7r and v7m.
2796 (RETLDM): Add Thumb-2 code.
2797 (do_it, shift1, do_push, do_pop, COND, THUMB_SYNTAX): New macros.
2798 * config/arm/arm.h (TARGET_CPU_CPP_BUILTINS): Define __thumb2__.
2799 (TARGET_THUMB1, TARGET_32BIT, TARGET_THUMB2, TARGET_DSP_MULTIPLY,
2800 TARGET_INT_SIMD, TARGET_UNIFIED_ASM, ARM_FT_STACKALIGN, IS_STACKALIGN,
2801 THUMB2_TRAMPOLINE_TEMPLATE, TRAMPOLINE_ADJUST_ADDRESS,
2802 ASM_OUTPUT_OPCODE, THUMB2_GO_IF_LEGITIMATE_ADDRESS,
2803 THUMB2_LEGITIMIZE_ADDRESS, CASE_VECTOR_PC_RELATIVE,
2804 CASE_VECTOR_SHORTEN_MODE, ADDR_VEC_ALIGN, ASM_OUTPUT_CASE_END,
2805 ADJUST_INSN_LENGTH): Define.
2806 (TARGET_REALLY_IWMMXT, TARGET_IWMMXT_ABI, CONDITIONAL_REGISTER_USAGE,
2807 STATIC_CHAIN_REGNUM, HARD_REGNO_NREGS, INDEX_REG_CLASS,
2808 BASE_REG_CLASS, MODE_BASE_REG_CLASS, SMALL_REGISTER_CLASSES,
2809 PREFERRED_RELOAD_CLASS, SECONDARY_OUTPUT_RELOAD_CLASS,
2810 SECONDARY_INPUT_RELOAD_CLASS, LIBCALL_VALUE, FUNCTION_VALUE_REGNO_P,
2811 TRAMPOLINE_SIZE, INITIALIZE_TRAMPOLINE, HAVE_PRE_INCREMENT,
2812 HAVE_POST_DECREMENT, HAVE_PRE_DECREMENT, HAVE_PRE_MODIFY_DISP,
2813 HAVE_POST_MODIFY_DISP, HAVE_PRE_MODIFY_REG, HAVE_POST_MODIFY_REG,
2814 REGNO_MODE_OK_FOR_BASE_P, LEGITIMATE_CONSTANT_P,
2815 REG_MODE_OK_FOR_BASE_P, REG_OK_FOR_INDEX_P, GO_IF_LEGITIMATE_ADDRESS,
2816 LEGITIMIZE_ADDRESS, THUMB2_LEGITIMIZE_ADDRESS,
2817 GO_IF_MODE_DEPENDENT_ADDRESS, MEMORY_MOVE_COST, BRANCH_COST,
2818 ASM_APP_OFF, ASM_OUTPUT_CASE_LABEL, ARM_DECLARE_FUNCTION_NAME,
2819 FINAL_PRESCAN_INSN, PRINT_OPERAND_PUNCT_VALID_P,
2820 PRINT_OPERAND_ADDRESS): Adjust for Thumb-2.
2821 (arm_arch_notm, arm_arch_thumb2, arm_arch_hwdiv): New declarations.
2822 * config/arm/arm-cores.def: Add arm1156t2-s, cortex-a8, cortex-r4 and
2824 * config/arm/arm-tune.md: Regenerate.
2825 * config/arm/arm-protos.h: Update prototypes.
2826 * config/arm/vfp.md: Enable patterns for Thumb-2.
2827 (arm_movsi_vfp): Add movw alternative. Use output_move_vfp.
2828 (arm_movdi_vfp, movsf_vfp, movdf_vfp): Use output_move_vfp.
2829 (thumb2_movsi_vfp, thumb2_movdi_vfp, thumb2_movsf_vfp,
2830 thumb2_movdf_vfp, thumb2_movsfcc_vfp, thumb2_movdfcc_vfp): New.
2831 * config/arm/libunwind.S: Add Thumb-2 code.
2832 * config/arm/constraints.md: Update include Thumb-2.
2833 * config/arm/ieee754-sf.S: Add Thumb-2/Unified asm support.
2834 * config/arm/ieee754-df.S: Ditto.
2835 * config/arm/bpabi.S: Ditto.
2836 * config/arm/t-arm (MD_INCLUDES): Add thumb2.md.
2837 * config/arm/predicates.md (low_register_operand,
2838 low_reg_or_int_operand, thumb_16bit_operator): New.
2839 (thumb_cmp_operand, thumb_cmpneg_operand): Rename...
2840 (thumb1_cmp_operand, thumb1_cmpneg_operand): ... to this.
2841 * config/arm/t-arm-elf: Add armv7 multilib.
2842 * config/arm/arm.md: Update patterns for Thumb-2 and Unified asm.
2844 (UNSPEC_STACK_ALIGN, ce_count): New.
2845 (arm_incscc, arm_decscc, arm_umaxsi3, arm_uminsi3,
2846 arm_zero_extendsidi2, arm_zero_extendqidi2): New
2848 * config/arm/fpa.md: Update patterns for Thumb-2 and Unified asm.
2849 (thumb2_movsf_fpa, thumb2_movdf_fpa, thumb2_movxf_fpa,
2850 thumb2_movsfcc_fpa, thumb2_movdfcc_fpa): New insns.
2851 * config/arm/cirrus.md: Update patterns for Thumb-2 and Unified asm.
2852 (cirrus_thumb2_movdi, cirrus_thumb2_movsi_insn,
2853 thumb2_cirrus_movsf_hard_insn, thumb2_cirrus_movdf_hard_insn): New
2855 * doc/extend.texi: Document ARMv7-M interrupt functions.
2856 * doc/invoke.texi: Document Thumb-2 new cores+architectures.
2858 2007-01-03 Jakub Jelinek <jakub@redhat.com>
2860 * unwind-dw2.c (SIGNAL_FRAME_BIT, EXTENDED_CONTEXT_BIT): Define.
2861 (struct _Unwind_Context): Rename args_size to flags, remove
2862 signal_frame field, add a new args_size field and version field.
2863 (_Unwind_IsSignalFrame, _Unwind_SetSignalFrame,
2864 _Unwind_IsExtendedContext): New inline functions.
2865 (_Unwind_GetGR, _Unwind_SetGR, _Unwind_GetGRPtr, _Unwind_SetGRPtr):
2866 Assume by_value array is only present if _Unwind_IsExtendedContext.
2867 (_Unwind_GetIPInfo, execute_cfa_program, uw_frame_state_for): Use
2868 _Unwind_IsSignalFrame.
2869 (__frame_state_for): Initialize context.flags to EXTENDED_CONTEXT_BIT.
2870 (uw_update_context_1): Use _Unwind_SetSignalFrame.
2871 (uw_init_context_1): Initialize context->flags to
2872 EXTENDED_CONTEXT_BIT.
2873 * config/rs6000/linux-unwind.h (frob_update_context): Use
2874 _Unwind_SetSignalFrame.
2876 2007-01-03 Andrew Pinski <andrew_pinski@playstation.sony.com>
2879 * gimplify.c (gimplify_modify_expr_complex_part): Move below
2880 tree_to_gimple_tuple. Call tree_to_gimple_tuple when we need
2883 2007-01-03 Kazu Hirata <kazu@codesourcery.com>
2885 * config/i386/i386.h (NON_STACK_REG_P, REGNO_OK_FOR_SIREG_P,
2886 REGNO_OK_FOR_DIREG_P, REWRITE_ADDRESS, ASM_OPERAND_LETTER,
2887 RET, AT_SP): Remove.
2888 * config/i386/i386.md (*sse_prologue_save_insn): Use return
2891 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
2892 * cfgbuild.c (find_basic_blocks): Likewise.
2893 * cfgrtl.c (rtl_create_basic_block): Likewise.
2894 * function.c (temp_slots_at_level): Likewise.
2895 * reg-stack.c (stack_regs_mentioned): Likewise.
2896 * regclass.c (allocate_reg_info): Likewise.
2897 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
2898 set_bb_for_stmt, move_block_to_fn): Likewise.
2899 * tree-complex.c (tree_lower_complex): Likewise.
2900 * vec.h (VEC_safe_grow_cleared): New.
2902 * cgraphunit.c, tree-ssa-alias.c: Fix comment typos.
2904 2007-01-03 Zdenek Dvorak <dvorakz@suse.cz>
2906 * loop-unswitch.c (unswitch_loop): Pass probabilities to loopify.
2907 * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Pass probabilities
2909 * cfgloopmanip.c (scale_loop_frequencies): Export.
2910 (loopify): Scale the frequencies by prescribed coefficients.
2911 (set_zero_probability): New function.
2912 (duplicate_loop_to_header_edge): Improve updating of frequencies.
2913 (lv_adjust_loop_entry_edge, loop_version): Set probabilities
2914 and frequencies according to arguments.
2915 * tree-ssa-loop-manip.c (tree_unroll_loop): Set probabilities
2917 * cfg.c (scale_bbs_frequencies_int): Allow scaling the frequencies up.
2918 * modulo-sched.c (sms_schedule): Set probabilities for entering
2919 versioned loop correctly.
2920 * tree-vect-transform.c (vect_transform_loop): Ditto.
2921 * cfgloop.h (loopify, loop_version): Declaration changed.
2922 (scale_loop_frequencies): Declared.
2924 2007-01-02 Jan Hubicka <jh@suse.cz>
2926 * cgraph.c: Include tree-flow.h
2927 (cgraph_add_new-function): Handle IPA_SSA mode; execute
2929 * cgraph.h (enum cgraph_state): Add CGRAPH_STATE_IPA_SSA.
2930 * tree-pass.h (pass_all_early_optimizations): Declare.
2931 * cgraphunit.c (cgraph_process_new_functions): Add IPA_SSA; execute
2933 (cgraph_analyze_function): Do early_local_passes.
2934 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2935 Do not add referenced vars.
2936 * tree-optimize.c (gate_all_optimizations): Do not execute when not in
2938 (gate_all_early_local_passes): New gate.
2939 (pass_early_local_passes): Use new gate.
2940 (execute_early_local_optimizations): New functions.
2941 (gate_all_early_optimizations): New gate.
2942 (pass_all_early_optimizations): New pass.
2943 (execute_free_datastructures): Free SSA only when initialized.
2944 (gate_init_datastructures): Init only when optimizing.
2945 (tree_lowering_passes): Do early local passes when called late.
2946 * tree-profile.c (do_tree_profiling): Don't profile functions added
2948 (do_early_tree_profiling, pass_early_tree_profile): Kill.
2949 * tree-cfg.c (update_modified_stmts): Do not update when operands are
2951 * passes.c (init_optimizations_passes): Reorder so we go into SSA
2952 during early_local_passes.
2953 * Makefile.in (cgraph.o): Add dependency on tree-flow.h.
2956 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2958 * Makefile.in: Update copyright year.
2960 2007-01-02 Carlos O'Donell <carlos@codesourcery.com>
2962 * Makefile.in: Export GCC_EXEC_PREFIX before calling $(RUNTEST)
2963 in $(lang_checks) and check-consistency targets.
2965 2007-01-02 Jan Hubicka <jh@suse.cz>
2967 * tree-mudflap.c (mf_decl_cache_locals, mf_build_check_statement_for):
2968 Do not add referenced vars.
2969 * tree-cfg.c (update_modified_stmts): Do not update when SSA operands
2971 * passes.c (init_optimization_passes): Put mudflap_2 after
2972 free_datastructures.
2974 2007-01-02 Jan Hubicka <jh@suse.cz>
2976 * tree-optimize (execute_fixup_cfg): Set after_inlining flag.
2977 Set NOTHROW flag on call statements proved to be nothrow.
2978 Update statement of local calls so new pure/const functions are
2979 updated. Update_ssa when in ssa form. Mark PHI nodes of nonlocal
2981 (tree_rest_of_compilation): Register hooks and initialize bitmap
2982 early. Do not set after_inlining flag.
2984 2007-01-02 Steve Ellcey <sje@cup.hp.com>
2986 * sbitmap.c (HOST_BITS_PER_LONG_LONG): Change to
2987 HOST_BITS_PER_LONGLONG
2989 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2992 * c-typeck.c (store_init_value): Don't emit pedantic overflow
2993 warning for non-static initializers.
2995 2007-01-02 Steven Bosscher <steven@gcc.gnu.org>
2997 * config/alpha/alpha.md, arm/arm.c, darwin.c, frv/frv.md,
2998 m32r/m32r.c, m32r/m32r.c, mn10300/mn10300.md, pa/pa.c,
2999 rs6000/rs6000.c, s390/s390.md, sh/sh.md, sparc/sparc.c:
3000 Always use set_unique_reg_note to add REG_EQUAL notes.
3002 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3005 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3007 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3008 * cfgbuild.c (find_basic_blocks): Likewise.
3009 * cfgrtl.c (rtl_create_basic_block): Likewise.
3010 * function.c (temp_slots_at_level): Likewise.
3011 * reg-stack.c (stack_regs_mentioned): Likewise.
3012 * regclass.c (allocate_reg_info): Likewise.
3013 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3014 set_bb_for_stmt, move_block_to_fn): Likewise.
3015 * tree-complex.c (tree_lower_complex): Likewise.
3016 * vec.h (VEC_safe_grow_cleared): New.
3018 2007-01-02 Ian Lance Taylor <iant@google.com>
3020 * c-common.c (c_common_truthvalue_conversion): When warning about
3021 using an assignment as a truth value, set TREE_NO_WARNING.
3023 2007-01-02 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3026 * c.opt (Wold-style-declaration): New.
3027 * doc/invoke.texi (C-only Warning Options): New.
3028 (Wold-style-declaration): Document it.
3029 (Wextra): Enabled by -Wextra.
3030 * c-opts.c (c_common_post_options): Enabled by -Wextra.
3031 * c-decl.c (declspecs_add_scspec): Replace -Wextra with
3032 -Wold-style-declaration.
3034 2007-01-02 Kazu Hirata <kazu@codesourcery.com>
3036 * alias.c (init_alias_analysis): Use VEC_safe_grow_cleared.
3037 * cfgbuild.c (find_basic_blocks): Likewise.
3038 * cfgrtl.c (rtl_create_basic_block): Likewise.
3039 * function.c (temp_slots_at_level): Likewise.
3040 * reg-stack.c (stack_regs_mentioned): Likewise.
3041 * regclass.c (allocate_reg_info): Likewise.
3042 * tree-cfg.c (init_empty_tree_cfg, build_tree_cfg, create_bb,
3043 set_bb_for_stmt, move_block_to_fn): Likewise.
3044 * tree-complex.c (tree_lower_complex): Likewise.
3045 * vec.h (VEC_safe_grow_cleared): New.
3047 2007-01-02 Douglas Gregor <doug.gregor@gmail.com>
3049 * c-common.c(c_common_nodes_and_builtins): Since variants of
3050 void_type_node get built before it is given a name, we need to
3051 give those variants the name, too.
3052 (complete_array_type): We need to work with the canonical main
3053 type of the array, from which we will build the qualified version.
3054 * params.def (PARAM_VERIFY_CANONICAL_TYPES): New.
3055 * print-tree.c (print_node): Display canonical type information
3057 * stor-layout.c (layout_type): When we don't know the
3058 alignment of a type for which we're building an array, we end up
3059 guessing wrong, so make the type require structural equality.
3060 * tree.c (make_node_stat): When we build a new type, it is its
3062 (build_type_attribute_qual_variant): When building an attribute
3063 variant, its canonical type is the non-attribute variant. However,
3064 if the attributes are target-dependent and they differ, we need to
3065 use structural equality checks for this type.
3066 (build_qualified_type): A qualified type is not equivalent to its
3067 unqualified variant; set the canonical type appropriately.
3068 (build_distinct_type_copy): When building a distinct type from
3069 another type, the new type is its own canonical type.
3070 (build_variant_type_copy): When building a new type variant, we
3071 assume that it is equivalent to the original type.
3072 (build_pointer_type_for_mode): When building a pointer type, also
3073 build a canonical type pointer.
3074 (build_reference_type_for_mode): When building a reference type,
3075 also build a canonical type reference.
3076 (build_index_type): When we can't hash an index type (e.g.,
3077 because its maximum value is negative), the index type requires
3078 structural equality tests.
3079 (build_array_type): Build the canonical form of an array type.
3080 (build_function_type): Function types require structural equality,
3081 because they contain default arguments, attributes, etc.
3082 (build_method_type_directly): Ditto for method types.
3083 (build_offset_type): Build the canonical offset type.
3084 (build_complex_type): Build the canonical vector type.
3085 (make_vector_type): Build the canonical vector type.
3086 * tree.h (TYPE_CANONICAL): New.
3087 (TYPE_STRUCTURAL_EQUALITY_P): New.
3088 (SET_TYPE_STRUCTURAL_EQUALITY): New.
3089 (struct tree_type): Added "canonical" field.
3090 * params.h (VERIFY_CANONICAL_TYPES): New.
3091 * doc/c-tree.texi (TYPE_CANONICAL): Document.
3092 (TYPE_STRUCTURAL_EQUALITY_P): Document.
3093 (SET_TYPE_STRUCTURAL_EQUALITY): Document.
3094 * doc/invoke.texi (verify-canonical-types): Document --param
3095 parameter for verifying canonical types.
3097 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3099 * config.gcc (powerpc-*-eabispe*, powerpc-*-eabisimaltivec*,
3100 powerpc-*-eabisim*, powerpc-*-eabialtivec*, powerpc-*-eabi*,
3101 powerpc-*-rtems*, powerpc-wrs-vxworks, powerpc-wrs-vxworksae,
3102 powerpcle-*-eabisim*, powerpcle-*-eabi*): Add rs6000/e500.h to
3104 * config/rs6000/e500.h: New.
3105 * config/rs6000/eabi.h (TARGET_SPE_ABI, TARGET_SPE, TARGET_E500,
3106 TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE, TARGET_E500_DOUBLE):
3108 * config/rs6000/linuxspe.h (TARGET_SPE_ABI, TARGET_SPE,
3109 TARGET_E500, TARGET_ISEL, TARGET_FPRS, TARGET_E500_SINGLE,
3110 TARGET_E500_DOUBLE): Remove.
3111 * config/rs6000/vxworks.h (TARGET_SPE_ABI, TARGET_SPE,
3112 TARGET_E500, TARGET_ISEL, TARGET_FPRS): Remove.
3113 * config/rs6000/rs6000.h (CHECK_E500_OPTIONS): Define.
3114 * config/rs6000/rs6000.c (rs6000_override_options): Use
3117 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3119 * config/rs6000/rs6000.c (print_operand): Check (TARGET_SPE ||
3120 TARGET_E500_DOUBLE), not TARGET_E500, for %y.
3121 (rs6000_generate_compare, rs6000_emit_sCOND, output_cbranch,
3122 rs6000_emit_cmove): Don't check TARGET_E500.
3123 * config/rs6000/rs6000.md (bunordered, bordered, sunordered,
3124 sordered): Don't check TARGET_E500.
3126 2007-01-01 Eric Christopher <echristo@apple.com>
3128 * config/mips/mips.c (mips_regno_mode_ok_for_base_p): Use
3129 HARD_REGISTER_NUM_P.
3131 2007-01-01 Roger Sayle <roger@eyesopen.com>
3133 * fold-const.c (fold_binary) <EQ_EXPR>: Fold "(X^C1) eq/ne C2" into
3134 "X eq/ne (C1^C2)". Fold "(X^Z) eq/ne (Y^Z)" as "X eq/ne Y" when Z
3135 has no side-effects. Fold "(X^C1) eq/ne (Y^C2)" as "(X^(C1^C2))
3138 2007-01-01 Mike Stump <mrs@apple.com>
3140 * configure.ac: Remove support for building with Apple's gcc-3.1.
3142 2007-01-02 Joseph Myers <joseph@codesourcery.com>
3145 * caller-save.c (add_stored_regs): Only handle SUBREGs if inner
3146 REG is a hard register. Do not modify REG before calling
3148 * rtlanal.c (subreg_get_info): Don't assert size of XMODE is a
3149 multiple of the size of YMODE for certain lowpart cases.
3151 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3154 * gimplify (voidify_wrapper_expr): Update for
3157 2007-01-01 Andreas Schwab <schwab@suse.de>
3160 * config/ia64/ia64.c (ia64_compute_frame_size): Account space for
3161 save of BR0 in extra_spill_size instead of spill_size.
3162 (ia64_expand_prologue): Save BR0 outside of the gr/br/fr spill
3164 (ia64_expand_epilogue): Restore BR0 from its new location.
3166 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3168 * gimplify.c (gimplify_init_constructor <case VECTOR_TYPE>):
3169 Use a temporary variable if the left hand side is not a gimple
3172 2007-01-01 Andrew Pinski <pinskia@gmail.com>
3174 * gimplify.c (gimplify_return_expr): Make the temporary variable
3175 for the return expression, a gimple register variable.
3177 2007-01-01 Jan Hubicka <jh@suse.cz>
3179 * emit-rtl.c (emit_copy_of_insn_after): Do not call copy_insn_1 for
3182 2007-01-01 Mike Stump <mrs@apple.com>
3184 * configure.ac (HAVE_GAS_LITERAL16): Add autoconf check for
3186 * config/darwin.c (machopic_select_rtx_section): Use
3188 (darwin_mergeable_constant_section): Likewise.
3189 * configure: Regenerate.
3190 * config.in: Regenerate.
3192 2007-01-01 Jan Hubicka <jh@suse.cz>
3193 Andrew Pinski <pinskia@gmail.com>
3195 * cgraphunit.c (cgraph_optimize): Call cgraph_add_new_functions
3196 before starting IPA passes.