Uros Bizjak [Wed, 6 Dec 2006 14:56:10 +0000 (15:56 +0100)]
i386.md (DIRFLAG_REG): Remove constant.
* config/i386/i386.md (DIRFLAG_REG): Remove constant.
(type): Remove cld.
(length_immediate): Do not depend on cld type attribute.
(length_address): Ditto.
(modrm): Ditto.
(memory): Ditto.
(cld): Remove insn pattern.
(strmov_singleop): Do not use DIRFLAG_REG.
(*strmovdi_rex_1): Ditto.
(*strmovsi_1): Ditto.
(*strmovsi_rex_1): Ditto.
(*strmovhi_1): Ditto.
(*strmovhi_rex_1): Ditto.
(*strmovqi_1): Ditto.
(*strmovqi_rex_1): Ditto.
(rep_mov): Ditto.
(*rep_movdi_rex64): Ditto.
(*rep_movsi): Ditto.
(*rep_movsi_rex64): Ditto.
(*rep_movqi): Ditto.
(*rep_movqi_rex64): Ditto.
(strset_singleop): Ditto.
(*strsetdi_rex_1): Ditto.
(*strsetsi_1): Ditto.
(*strsetsi_rex_1): Ditto.
(*strsethi_1): Ditto.
(*strsethi_rex_1): Ditto.
(*strsetqi_1): Ditto.
(*strsetqi_rex_1): Ditto.
(rep_stos): Ditto.
(*rep_stosdi_rex64): Ditto.
(*rep_stossi): Ditto.
(*rep_stossi_rex64): Ditto.
(*rep_stosqi): Ditto.
(*rep_stosqi_rex64): Ditto.
(cmpstrnsi): Do not generate cld insn.
(cmpstrnqi_nz_1): Do not use DIRFLAG_REG.
(*cmpstrnqi_nz_1): Ditto.
(*cmpstrnqi_nz_rex_1): Ditto.
(cmpstrnqi_1): Ditto.
(*cmpstrnqi_1): Ditto.
(*cmpstrnqi_rex_1): Ditto.
(strlenqi_1): Ditto.
(*strlenqi_1): Ditto.
(*strlenqi_rex_1): Ditto.
* config/i386/geode.md (shift): Do not depend on cld type attribute.
* config/i386/pentium.md (pent_cld): Remove insn reservation.
* config/i386/athlon.md (athlon_decode): Do not depend on cld
type attribute.
* config/i386/ppro.md (ppro_cld): Remove insn reservation.
* config/i386/k6.md (k6_alux_only): Do not depend on cld type
attribute.
(k6_alux_only_load): Ditto.
(k6_alux_only_store): Ditto.
* config/i386/i386.c (ix86_expand_movmem): Remove cld_done boolean.
Do not emit cld instruction.
(ix86_expand_setmem): Ditto.
(ix86_expand_strlen): Do not emit cld instruction.
Mark Mitchell [Wed, 6 Dec 2006 05:12:46 +0000 (05:12 +0000)]
re PR c++/29729 (ICE with template class in template function)
PR c++/29729
* decl2.c (check_member_template): Move check for member
templates in local classes to ...
* parser.c (cp_parser_template_declaration_after_export):
... here.
PR c++/29729
* g++.dg/template/crash63.C: New test.
Bernd Schmidt [Wed, 6 Dec 2006 00:25:29 +0000 (00:25 +0000)]
reload1.c (delete_output_reload): Count occurrences in CALL_INSN_FUNCTION_USAGE.
* reload1.c (delete_output_reload): Count occurrences in
CALL_INSN_FUNCTION_USAGE.
* rtlanal.c (count_occurrences): Handle EXPR_LIST nodes without
crashing at the end of the list.
Mark Mitchell [Tue, 5 Dec 2006 21:33:20 +0000 (21:33 +0000)]
re PR c++/29728 (ICE on invalid initializer in template function)
PR c++/29728
* decl.c (check_array_designated_initializer): New function.
(maybe_deduce_size_from_array_init): Use it.
(reshape_init_array): Likewise.
PR c++/29728
* g++.dg/template/crash62.C: New test.
Paul Thomas [Tue, 5 Dec 2006 19:45:25 +0000 (19:45 +0000)]
re PR fortran/30003 ([4.1 only] Expressions with side effects in array references)
2006-12-05 Paul Thomas <pault@gcc.gnu.org>
PR fortran/30003
* trans-array.c (gfc_trans_create_temp_array): Set the section
ends to zero.
(gfc_conv_array_transpose): Likewise.
(gfc_conv_section_startstride): Declare an expression for end,
set it from a the array reference and evaluate it for the info
structure. Zero the ends in the ss structure and set end, used
in the bounds check, from the info structure.
trans.h: Add and end array to the gfc_ss_info structure.
2006-12-05 Paul Thomas <pault@gcc.gnu.org>
PR fortran/30003
* gfortran.dg/allocatable_function_1.f90: Increase the number
of expected calls of free to 10; the lhs section reference is
now evaluated so there is another call to bar. Change the
comment appropriately.
* gfortran.dg/array_section_1.f90: New test.
PR fortran/29912
* trans-expr.c (gfc_trans_arrayfunc_assign): Return NULL if the
lhs and rhs character lengths are not constant and equal for
character array valued functions.
2006-12-05 Paul Thomas <pault@gcc.gnu.org>
PR fortran/29912
* gfortran.dg/char_result_12.f90: New test.
Paolo Carlini [Tue, 5 Dec 2006 19:25:48 +0000 (19:25 +0000)]
DR 543, [Ready].
2006-12-05 Paolo Carlini <pcarlini@suse.de>
DR 543, [Ready].
* include/bits/slice_array.h (slice::slice()): Implement the
resolution.
* include/bits/gslice.h (gslice::_Indexer::_Indexer()): Add.
(gslice::gslice()): Use it.
* testsuite/26_numerics/valarray/dr543.cc: New.
* docs/html/ext/howto.html: Add an entry for DR 543.
* config/i386/i386.c (ix86_builtin_vectorized_function): Declare.
(TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Define.
(ix86_builtin_vectorized_function): New function to vectorize
sqrt.
* gcc.dg/vect/vect.exp: Add support for -fno-math-errno tests.
* gcc.dg/vect/vect-pow-2.c: Rename to ...
* gcc.dg/vect/no-math-errno-vect-pow-1.c: ... this. Require
vect_double, xfail for spu*-*-*.
Bernd Schmidt [Tue, 5 Dec 2006 08:49:56 +0000 (08:49 +0000)]
rtlanal.c (note_uses): Deal with SEQUENCEs.
* rtlanal.c (note_uses): Deal with SEQUENCEs.
* config/bfin/bfin.c: Include "timevar.h".
(bfin_flag_schedule_insns2, splitting_for_sched,
bfin_flag_var_tracking): New variables.
(print_operand): Handle '%!'.
(override_options): Disable normal sched2 pass, instead set
bfin_flag_schedule_insns2 for reorg to handle it.
(output_file_start): Likewise for var-tracking.
(bfin_optimize_loop): Take some care not to stumble over SEQUENCEs.
(gen_one_bundle, bfin_gen_bundles, type_for_anomaly, trapping_loads_p):
New functions.
(bfin_reorg): Do second scheduling pass here, and call
bfin_gen_bundles. Use type_for_anomaly and trapping_loads_p instead
of examining insns directly. If bfin_flag_var_tracking, call
var-tracking pass when done with everything else.
* config/bfin/bfin.h (PRINT_OPERAND_PUNCT_VALID_P): '!' is valid.
* config/bfin/bfin.md (UNSPEC_32BIT): New constant.
(movsi_insn32): New pattern, with two new splits to create it
before the final scheduling pass.
(neghi2): Not a dsp32 insn, rather alu0.
(movbi, pushsi_insn, popsi_insn, movsi_insn, movv2hi_insn, movhi_insn,
movqi_insn, movsf_insn, movsi_insv, extendhisi2, zero_extendhisi2,
extendqihi2, extendqisi2, zero_extendqihi2, zero_extendqisi2,
mulhisi3, umulhisi3, ssadsi3, sssubsi3, smaxsi3, sminsi3, abssi2,
ssnegsi2, signbitssi2, smaxhi3, sminhi3, abshi2, ssneghi2, signbitshi2,
movhi_low2high, movhi_high2high, movhi_low2low, movhi_high2low,
movhiv2hi_low, movhiv2hi_high, composev2hi, packv2hi, movv2hi_hi,
ssaddhi3, sssubhi3, addv2hi3, ssaddv2hi3, subv2hi3, sssubv2hi3,
addsubv2hi3, subaddv2hi3, ssaddsubv2hi3, sssubaddv2hi3, sublohiv2hi3,
subhilov2hi3, sssublohiv2hi3, sssubhilov2hi3, addlohiv2hi3,
addhilov2hi3, ssaddlohiv2hi3, ssaddhilov2hi3, sminv2hi3, smaxv2hi3,
flag_mulhi, flag_mulhisi, flag_mulhisi_parts, flag_machi,
flag_machi_acconly, flag_macinithi, flag_macinit1hi, mulv2hi3,
flag_mulv2hi, flag_mulv2hi_parts, flag_macv2hi_parts,
flag_macv2hi_parts_acconly, flag_macinitv2hi_parts,
flag_macinit1v2hi_parts, mulhisi_ll, mulhisi_lh, mulhisi_hl,
mulhisi_hh, ssnegv2hi2, absv2hi2, ssashiftv2hi3, ssashifthi3,
lshiftv2hi3, lshifthi3): Use '%!' to terminate all dsp32/load/store
variants instead of ';'.
(ror_one, rol_one): Likewise. Make them dsp32 insns.
(ashlsi3_insn, ashrsi3, lshrsi3): Add dsp32 variants.
(align8, align16, align24): Now named patterns; also using '%!'.
(mnop): New insn.
PR fortran/29962
* initialization_4.f90: Test noninteger exponents (-std=f95).
* initialization_5.f90: New test for noninteger exponents with -std=f2003
Daniel Berlin [Mon, 4 Dec 2006 19:07:05 +0000 (19:07 +0000)]
tree-ssa-alias.c (compute_may_aliases): Compute flow sensitive before flow insensitive.
2006-12-04 Daniel Berlin <dberlin@dberlin.org>
* tree-ssa-alias.c (compute_may_aliases):
Compute flow sensitive before flow insensitive.
(compute_flow_sensitive_aliasing): Don't try to add
a symbol memory tag to itself.
* alias.h (alias_set_subset_of): New prototype.
* alias.c (alias_set_subset_of): New function.
* tree-ssa-structalias.c: Update comments.
(struct variable_info): Add finished_solution.
(new_varinfo): Set finished solution to NULL.
(var_escaped_vars): Remove.
(escaped_vars_tree): Remove.
(escaped_vars_id): Remove.
(nonlocal_vars_id): Remove.
(constraint_expr_type): Add INCLUDES.
(graph_size): Removed.
(dump_constraint): Support INCLUDES.
(build_constraint_graph): Ditto.
(collapse_nodes): Add merge_solutions argument.
Don't merge attributes.
(process_unification_queue): Just use collapse_nodes.
(perform_var_substitution): Update call to collapse_nodes.
(get_constraint_exp_from_ssa_var): Use INCLUDES.
(process_constraint): Fix non-field sensitive handling
Handle includes.
(get_constraint_for): Use INCLUDES.
(make_constraint_from_anything): Renamed from
make_constraint_from_escaped.
(make_constraint_to_escaped): Removed.
(find_global_initializers): Removed.
(create_variable_info_for): Do not make constraints to escaped
vars anymore.
(dump_solution_for_var): Don't print out the equivalent points-to
sets, just use the name of the variable it shares it with.
(intra_create_variable_infos): Use INCLUDES.
Change nonlocal variable sets to anything sets.
(init_base_vars): Remove escaped_vars and nonlocal_var
initialization.
(find_escape_constraints): Removed.
(delete_points_to_sets): Remove dead code.
(used_smt_calculated): New variable.
(set_used_smts): New function.
(merge_smts_into): New function.
(find_what_p_points_to): Modify to use SMTs.
(create_nonlocal_var): Remove.
* tree-ssa-operands.c (access_can_touch_variable): Remove
reference to nonlocal_all.
Mark Mitchell [Mon, 4 Dec 2006 18:00:32 +0000 (18:00 +0000)]
re PR c++/29632 (ICE on invalid code: regenerate_decl_from_template, at cp/pt.c:10969)
PR c++/29632
* call.c (add_builtin_candidate): Do not permit NULL pointer
constants to be compared with template parameters.
PR c++/29632
* g++.dg/template/error23.C: New test.
Andrew MacLeod [Mon, 4 Dec 2006 14:26:37 +0000 (14:26 +0000)]
Switch live on entry to a per block basis from per variable.
* tree-outof-ssa.c (coalesce_ssa_name): Use calculate_live_ranges.
* tree-ssa-live.c (new_tree_live_info, delete_tree_live_info): Update.
(add_livein_if_notdef): Delete.
(loe_visit_block): New. Propogate live on entry info for a block into
each predecessor. If it changes, make sure it is visited again.
(live_worklist): Visit every block and update the live on entry info
for preds. Iterate over any that changed.
(set_var_live_on_entry): Populate the live on entry blocks with bits
based on the immediate uses of a var.
(calculate_live_on_entry): Remove.
(calculate_live_on_exit): Calculate live on exit based on the newly
oriented live on entry bits.
(calculate_live_ranges): Build live on entry and exit vectors.
(dump_live_info): Use new orientation of live on entry bitmaps.
(verify_live_on_entry): New. Split out verification code from old
calculate_live_on_entry routine.
* tree-ssa-live.h (struct tree_live_info_d): Add Working stack.
(live_entry_blocks): Rename to live_on_entry and return bitmap for a
basic_block instead of for a partition.
(live_merge_and_clear): Add asserts.
(make_live_on_entry): Set partition bit in basic block vector.
Jakub Jelinek [Mon, 4 Dec 2006 10:26:00 +0000 (11:26 +0100)]
re PR middle-end/29947 (OpenMP parallel for fails for reversed loop range)
PR libgomp/29947
* omp-low.c (expand_omp_for_static_nochunk,
expand_omp_for_static_chunk): Do all arithmetics in signed rather than
unsigned type.
* loop.c (gomp_loop_init): Make parameters signed. Set ws->end to
start if there shouldn't be any loop iterations.
(gomp_loop_ordered_static_start): Remove start == end test.
* testsuite/libgomp.c/pr29947-1.c: New test.
* testsuite/libgomp.c/pr29947-2.c: New test.
Jakub Jelinek [Mon, 4 Dec 2006 10:20:51 +0000 (11:20 +0100)]
re PR middle-end/29965 (OpenMP vs always throw in a loop)
PR middle-end/29965
* omp-low.c (determine_parallel_type): Don't try to optimize combined
parallels if region->exit or region->inner->exit is NULL.
(expand_omp_parallel): Handle NULL region->exit even if cfg is already
built.
(expand_omp_for_generic): Handle NULL region->cond and region->exit.
(expand_omp_for): Use expand_omp_for_generic if either region->cond
or region->exit is NULL.
(expand_omp_sections): Handle NULL region->cond or region->exit.
(expand_omp_synch):
* gcc.dg/gomp/pr29965-1.c: New test.
* gcc.dg/gomp/pr29965-2.c: New test.
* gcc.dg/gomp/pr29965-3.c: New test.
* gcc.dg/gomp/pr29965-4.c: New test.
* gcc.dg/gomp/pr29965-5.c: New test.
* gcc.dg/gomp/pr29965-6.c: New test.
* g++.dg/gomp/pr29965-1.C: New test.
* g++.dg/gomp/pr29965-2.C: New test.
* g++.dg/gomp/pr29965-3.C: New test.
* g++.dg/gomp/pr29965-4.C: New test.
* g++.dg/gomp/pr29965-5.C: New test.
* g++.dg/gomp/pr29965-6.C: New test.
* g++.dg/gomp/pr29965-7.C: New test.
* g++.dg/gomp/pr29965-8.C: New test.
* g++.dg/gomp/pr29965-9.C: New test.
* g++.dg/gomp/pr29965-10.C: New test.
Eric Botcazou [Mon, 4 Dec 2006 07:47:32 +0000 (07:47 +0000)]
tree.c (walk_type_fields): Do not handle TYPE_MIN_VALUE and TYPE_MAX_VALUE for scalar types here but...
* tree.c (walk_type_fields): Do not handle TYPE_MIN_VALUE and
TYPE_MAX_VALUE for scalar types here but...
(walk_tree): ...there instead. Return NULL_TREE if the TYPE_DECL
is attached an error mark, and the return value of the callback
if it is not NULL_TREE.
cp/
* pt.c (for_each_template_parm_r) <INTEGER_TYPE>: New case.
Call for_each_template_parm on TYPE_MIN_VALUE and TYPE_MAX_VALUE.
Dorit Nuzman [Sun, 3 Dec 2006 20:05:31 +0000 (20:05 +0000)]
vect.exp: Enable tests to run on spu.
* g++.dg/vect/vect.exp: Enable tests to run on spu.
* gcc.dg/vect/vect.exp: Enable tests to run on spu.
* gcc.dg/vect/no-section-anchors-vect-69.c: Reduce array size.
* gcc.dg/vect/vect-70.c: Reduce array size.
* gcc.dg/vect/fast-math-vect-reduc-7.c: Xfail for
vect_no_compaer_double targets.
* lib/target-supports.exp: Add spu to vect_no_align, no_vect_int_max,
vect_float, vect_double, and vect_int.
Paul Thomas [Sun, 3 Dec 2006 07:18:22 +0000 (07:18 +0000)]
re PR fortran/29642 (Fortran 2003: VALUE Attribute (call by value not call by reference for actual arguments))
2006-12-03 Paul Thomas <pault@gcc.gnu.org>
PR fortran/29642
* trans-expr.c (gfc_conv_variable): A character expression with
the VALUE attribute needs an address expression; otherwise all
other expressions with this attribute must not be dereferenced.
(gfc_conv_function_call): Pass expressions with the VALUE
attribute by value, using gfc_conv_expr.
* symbol.c (check_conflict): Add strings for INTENT OUT, INOUT
and VALUE. Apply all the constraints associated with the VALUE
attribute.
(gfc_add_value): New function.
(gfc_copy_attr): Call it for VALUE attribute.
* decl.c (match_attr_spec): Include the VALUE attribute.
(gfc_match_value): New function.
* dump-parse-tree.c (gfc_show_attr): Include VALUE.
* gfortran.h : Add value to the symbol_attribute structure and
add a prototype for gfc_add_value
* module.c (mio_internal_string): Include AB_VALUE in enum.
(attr_bits): Provide the VALUE string for it.
(mio_symbol_attribute): Read or apply the VLUE attribute.
* trans-types.c (gfc_sym_type): Variables with the VLAUE
attribute are not passed by reference!
* resolve.c (was_declared): Add value to those that return 1.
(resolve_symbol): Value attribute requires dummy attribute.
* match.h : Add prototype for gfc_match_public.
* parse.c (decode_statement): Try to match a VALUE statement.
2006-12-03 Paul Thomas <pault@gcc.gnu.org>
PR fortran/29642
* gfortran.dg/value_1.f90 : New test.
* gfortran.dg/value_2.f90 : New test.
* gfortran.dg/value_3.f90 : New test.
* gfortran.dg/value_4.f90 : New test.
* gfortran.dg/value_4.c : Called from value_4.f90.
Howard Hinnant [Sat, 2 Dec 2006 17:06:57 +0000 (17:06 +0000)]
acinclude.m4: Allow OPTIMIZE_CXXFLAGS to be set by configure.host.
2006-12-02 Howard Hinnant <hhinnant@apple.com>
* acinclude.m4: Allow OPTIMIZE_CXXFLAGS to be set by configure.host.
* configure.host: Set OPTIMIZE_CXXFLAGS to -fvisibility-inlines-hidden
for x86/darwin.
* configure: Regenerate.
* config/i386/i386.c (pentium4_cost, nocona_cost): Update
32bit memcpy/memset decriptors.
(decide_alg): With -minline-all-stringops and sizes that are best
to be copied via libcall still work hard enough to pick non-libcall
strategy.
* decl.c (poplevel): Check DECL_INITIAL invariant.
(duplicate_decls): Preserve DECL_INITIAL when eliminating
a new definition in favour of an old declaration.
(start_preparsed_function): Define and document value of
DECL_INITIAL before and after routine.
(finish_function): Check DECL_INITIAL invariant.
* parser.c
(cp_parser_function_definition_from_specifiers_and_declarator):
Skip duplicate function definitions.
Eric Botcazou [Fri, 1 Dec 2006 22:46:45 +0000 (22:46 +0000)]
fold-const.c (fold_binary): Use the precision of the type instead of the size of its mode to compute the...
* fold-const.c (fold_binary) <LT_EXPR>: Use the precision of the
type instead of the size of its mode to compute the highest and
lowest possible values. Still check the size of the mode before
flipping the signedness of the comparison.
Volker Reichelt [Fri, 1 Dec 2006 21:35:25 +0000 (21:35 +0000)]
re PR c++/30022 (ICE on vector operand in division)
PR c++/30022
* typeck.c (type_after_usual_arithmetic_conversions):
Fix assertion for vector types.
(build_binary_op): Use temporary for inner type of vector types.
Thomas Koenig [Fri, 1 Dec 2006 21:04:38 +0000 (21:04 +0000)]
re PR libfortran/29568 (implement unformatted files with subrecords (Intel style))
2006-12-01 Thomas Koenig <Thomas.Koenig@online.de>
PR libfortran/29568
* gfortran.dg/convert_implied_open.f90: Change to
new default record length.
* gfortran.dg/unf_short_record_1.f90: Adapt to
new error message.
* gfortran.dg/unformatted_subrecords_1.f90: New test.
2006-12-01 Thomas Koenig <Thomas.Koenig@online.de>
PR libfortran/29568
* gfortran.h (gfc_option_t): Add max_subrecord_length.
(top level): Define MAX_SUBRECORD_LENGTH.
* lang.opt: Add option -fmax-subrecord-length=.
* trans-decl.c: Add new function set_max_subrecord_length.
(gfc_generate_function_code): If we are within the main
program and max_subrecord_length has been set, call
set_max_subrecord_length.
* options.c (gfc_init_options): Add defaults for
max_subrecord_lenght, convert and record_marker.
(gfc_handle_option): Add handling for
-fmax_subrecord_length.
* invoke.texi: Document the new default for
-frecord-marker=<n>.
2006-12-01 Thomas Koenig <Thomas.Koenig@online.de>
PR libfortran/29568
* libgfortran/libgfortran.h (compile_options_t): Add
record_marker. (top level): Define GFC_MAX_SUBRECORD_LENGTH.
* runtime/compile_options.c (set_record_marker): Change
default to four-byte record marker.
(set_max_subrecord_length): New function.
* runtime/error.c (translate_error): Change error message
for short record on unformatted read.
* io/io.h (gfc_unit): Add recl_subrecord, bytes_left_subrecord
and continued.
* io/file_pos.c (unformatted_backspace): Change default of record
marker size to four bytes. Loop over subrecords.
* io/open.c: Default recl is max_offset. If
compile_options.max_subrecord_length has been set, set set
u->recl_subrecord to its value, to the maximum value otherwise.
* io/transfer.c (top level): Add prototypes for us_read, us_write,
next_record_r_unf and next_record_w_unf.
(read_block_direct): Separate codepaths for unformatted direct
and unformatted sequential. If a recl has been set by the
user, use the number of bytes left for the record if it is smaller
than the read request. Loop over subrecords. Set an error if the
user has set a recl and the read was short.
(write_buf): Separate codepaths for unformatted direct and
unformatted sequential. If a recl has been set by the
user, use the number of bytes left for the record if it is smaller
than the read request. Loop over subrecords. Set an error if the
user has set a recl and the read was short.
(us_read): Add parameter continued (to indicate that bytes_left
should not be intialized). Change default of record marker size
to four bytes. Use subrecord. If the subrecord length is smaller than
zero, this indicates a continuation.
(us_write): Add parameter continued (to indicate that the continued
flag should be set). Use subrecord.
(pre_position): Use 0 for continued on us_write and us_read calls.
(skip_record): New function.
(next_record_r_unf): New function.
(next_record_r): Use next_record_r_unf.
(write_us_marker): Default size for record markers is four bytes.
(next_record_w_unf): New function.
(next_record_w): Use next_record_w_unf.
Joseph Myers [Fri, 1 Dec 2006 02:25:22 +0000 (02:25 +0000)]
re PR target/24036 ([e500] ICE in subreg_offset_representable_p, at rtlanal.c:3143)
2006-12-01 Joseph Myers <joseph@codesourcery.com>
David Edelsohn <edelsohn@gnu.org>
PR target/24036
* doc/tm.texi (HARD_REGNO_NREGS_HAS_PADDING,
HARD_REGNO_NREGS_WITH_PADDING): Document new target macros.
* defaults.h (HARD_REGNO_NREGS_HAS_PADDING,
HARD_REGNO_NREGS_WITH_PADDING): Define.
* config/i386/i386.h (HARD_REGNO_NREGS_HAS_PADDING,
HARD_REGNO_NREGS_WITH_PADDING): Define.
* rtlanal.c (subreg_regno_offset, subreg_offset_representable_p):
Use new macros to detect modes with holes; do not look at integer
units.
(subreg_offset_representable_p): Check for and disallow cases
where the modes use different numbers of bits from registers.
* config/rs6000/rs6000.c (rs6000_emit_move): Handle TFmode
constant for soft-float.
(rs6000_hard_regno_nregs): Use UNITS_PER_FP_WORD for e500 GPRs
containing doubles.
(rs6000_split_multireg_move): Use DFmode reg_mode for TFmode moves
in E500 double case.
* config/rs6000/rs6000.md (movtf): Allow soft-float.
(movtf_softfloat): New.
Co-Authored-By: David Edelsohn <edelsohn@gnu.org>
From-SVN: r119395