Ian Lance Taylor [Thu, 22 Sep 2016 20:32:16 +0000 (20:32 +0000)]
compiler: compile runtime.getcaller{pc,sp} into builtin functions
The runtime functions runtime.getcallerpc and runtime.getcallersp are
intended to be efficient ways to get the return and frame address of the
caller (that is, the caller of runtime.getcallerpc). In the C code that
is implemented by simply using C macros:
This patch essentially implements those macros in the Go code.
It would be nice if we could just use //extern for this, but it doesn't
work because the runtime code passes the right argument. Of course we
could change the runtime code, but these are common enough that I'd
prefer to avoid the difference from the gc version of the runtime code.
This patch corrects the existing declaration of __builtin_return_address
to use uint32, rather than uint, for the parameter type. The builtin
functions take the C type "unsigned int", which for the targets we use
corresponds to the Go type uint32. Not that it should matter, really.
Martin Liska [Thu, 22 Sep 2016 12:52:09 +0000 (14:52 +0200)]
re PR ipa/77653 (wrong code at -Os and above on x86_64-linux-gnu (in both 32-bit and 64-bit modes))
Fix PR ipa/77653
PR ipa/77653
* gcc.dg/ipa/pr77653.c: New test.
PR ipa/77653
* ipa-icf.c (sem_variable::merge): Yield merge operation if
alias address matters, not necessarily address of original.
Richard Biener [Thu, 22 Sep 2016 12:16:12 +0000 (12:16 +0000)]
re PR tree-optimization/77677 (ICE at -O1 and above in both 32-bit and 64-bit modes on x86_64-linux-gnu (internal compiler error: in set_value_range, at tree-vrp.c:361))
2016-09-22 Richard Biener <rguenther@suse.de>
PR middle-end/77677
* gimple-match-head.c (gimple_resimplify1): Drop TREE_OVERFLOW
from constant folding results.
(gimple_resimplify2): Likewise.
(gimple_resimplify3): Likewise.
Paul Thomas [Thu, 22 Sep 2016 11:26:59 +0000 (11:26 +0000)]
re PR fortran/48298 ([F03] User-Defined Derived-Type IO (DTIO))
2016-09-22 Paul Thomas <pault@gcc.gnu.org>
PR fortran/48298
* gfortran.h : Place the pseudo operators INTRINSIC_FORMATTED
and INTRINSIC_UNFORMATTED after the sentinel in enum
gfc_intrinsic_op so that they do not appear as place holders
in module_write.
* interface.c (dtio_op): Comment on the special nature of the
pseudo operators INTRINSIC FORMATTED and INTRINSIC_UNFORMATTED.
Jakub Jelinek [Thu, 22 Sep 2016 11:17:24 +0000 (13:17 +0200)]
re PR tree-optimization/77665 (ICE in expand_GOMP_SIMD_VF, at internal-fn.c:172)
PR fortran/77665
* tree-inline.c (remap_gimple_stmt): Set has_simduid_loops
for all IFN_GOMP_SIMD_* internal fns, not just for
IFN_GOMP_SIMD_ORDERED_*.
Jonathan Wakely [Thu, 22 Sep 2016 10:06:41 +0000 (11:06 +0100)]
Update pretty printer for std::variant
* python/libstdcxx/v6/printers.py (StdVariantPrinter): Adjust for
recent change to _Variant_storage.
* testsuite/libstdc++-prettyprinters/cxx17.cc: Test variant with
reference type.
Paul Thomas [Thu, 22 Sep 2016 07:46:07 +0000 (07:46 +0000)]
interface.c (check_dtio_interface1): Introduce errors for alternate returns and incorrect numbers of arguments.
2016-09-22 Paul Thomas <pault@gcc.gnu.org>
* interface.c (check_dtio_interface1): Introduce errors for
alternate returns and incorrect numbers of arguments.
(gfc_find_specific_dtio_proc): Return cleanly if the derived
type either doesn't exist or has no namespace.
2016-09-22 Paul Thomas <pault@gcc.gnu.org>
* gfortran.dg/dtio_11.f90: Correct for changed error messages.
* gfortran.dg/dtio_13.f90: New test.
Tim Shen [Thu, 22 Sep 2016 03:15:58 +0000 (03:15 +0000)]
re PR libstdc++/77641 (std::variant copy-initialization fails for type with non-trivial constexpr ctor)
PR libstdc++/77641
* include/std/variant (_Variant_storage::_Variant_storage):
Change _Variant_storage's union to be default constructible.
* testsuite/20_util/variant/compile.cc: New test.
Michael Meissner [Wed, 21 Sep 2016 22:12:13 +0000 (22:12 +0000)]
re PR target/77670 (PowerPC64 Spec 2006 fails on 453.povray using -mcpu=power9 -mpower9-minmax)
2016-09-21 Michael Meissner <meissner@linux.vnet.ibm.com>
PR target/77670
* config/rs6000/predicates.md (invert_fpmask_comparison_operator):
New predicate that matches the ISA 3.0 XSCMP{EQ,GT,GE}DP
instructions when you want to invert the test.
* config/rs6000/rs6000.md (fpmask<mode>): Use the arguments in the
correct order for XXSEL.
(mov<SFDF:mode><SFDF2:mode>cc_invert_p9): Define the inverted test
for using XSCMP{EQ,GT,GE}DP.
Michael Meissner [Wed, 21 Sep 2016 22:11:42 +0000 (22:11 +0000)]
re PR target/77670 (PowerPC64 Spec 2006 fails on 453.povray using -mcpu=power9 -mpower9-minmax)
2016-09-21 Michael Meissner <meissner@linux.vnet.ibm.com>
PR target/77670
* config/rs6000/predicates.md (invert_fpmask_comparison_operator):
New predicate that matches the ISA 3.0 XSCMP{EQ,GT,GE}DP
instructions when you want to invert the test.
* config/rs6000/rs6000.md (fpmask<mode>): Use the arguments in the
correct order for XXSEL.
(mov<SFDF:mode><SFDF2:mode>cc_invert_p9): Define the inverted test
for using XSCMP{EQ,GT,GE}DP.
Ian Lance Taylor [Wed, 21 Sep 2016 20:58:51 +0000 (20:58 +0000)]
compiler, runtime: replace hashmap code with Go 1.7 hashmap
This change removes the gccgo-specific hashmap code and replaces it with
the hashmap code from the Go 1.7 runtime. The Go 1.7 hashmap code is
more efficient, does a better job on details like when to update a key,
and provides some support against denial-of-service attacks.
The compiler is changed to call the new hashmap functions instead of the
old ones.
The compiler now tracks which types are reflexive and which require
updating when used as a map key, and records the information in map type
descriptors.
Map_index_expression is simplified. The special case for a map index on
the right hand side of a tuple expression has been unnecessary for some
time, and is removed. The support for specially marking a map index as
an lvalue is removed, in favor of lowering an assignment to a map index
into a function call. The long-obsolete support for a map index of a
pointer to a map is removed.
The __go_new_map_big function (known to the compiler as
Runtime::MAKEMAPBIG) is no longer needed, as the new runtime.makemap
function takes an int64 hint argument.
The old map descriptor type and supporting expression is removed.
The compiler was still supporting the long-obsolete syntax `m[k] = 0,
false` to delete a value from a map. That is now removed, requiring a
change to one of the gccgo-specific tests.
The builtin len function applied to a map or channel p is now compiled
as `p == nil ? 0 : *(*int)(p)`. The __go_chan_len function (known to
the compiler as Runtime::CHAN_LEN) is removed.
Support for a shared zero value for maps to large value types is
introduced, along the lines of the gc compiler. The zero value is
handled as a common variable.
The hash function is changed to take a seed argument, changing the
runtime hash functions and the compiler-generated hash functions.
Unlike the gc compiler, both the hash and equal functions continue to
take the type length.
Types that can not be compared now store nil for the hash and equal
functions, rather than pointing to functions that throw. Interface hash
and comparison functions now check explicitly for nil. This matches the
gc compiler and permits a simple implementation for ismapkey.
The compiler is changed to permit marking struct and array types as
incomparable, meaning that they have no hash or equal function. We use
this for thunk types, removing the existing special code to avoid
generating hash/equal functions for them.
The C runtime code adds memclr, memequal, and memmove functions.
The hashmap code uses go:linkname comments to make the functions
visible, as otherwise the compiler would discard them.
The hashmap code comments out the unused reference to the address of the
first parameter in the race code, as otherwise the compiler thinks that
the parameter escapes and copies it onto the heap. This is probably not
needed when we enable escape analysis.
Several runtime map tests that ere previously skipped for gccgo are now
run.
The Go runtime picks up type kind information and stubs. The type kind
information causes the generated runtime header file to define some
constants, including `empty`, and the C code is adjusted accordingly.
A Go-callable version of runtime.throw, that takes a Go string, is
added to be called from the hashmap code.
David Malcolm [Wed, 21 Sep 2016 20:55:06 +0000 (20:55 +0000)]
Introduce class rtx_reader
Bundle up various global variables within gensupport.c into a
class rtx_reader, with a view towards making it easier to run the
code more than once in-process.
gcc/ChangeLog:
* genconstants.c (main): Introduce noop_reader and convert call
to read_md_files to a method call.
* genenums.c (main): Likewise.
* genmddeps.c (main): Likewise.
* genpreds.c (write_tm_constrs_h): Replace use of "in_fname" with
rtx_reader_ptr->get_top_level_filename ().
(write_tm_preds_h): Likewise.
(write_insn_preds_c): Likewise.
* gensupport.c (class gen_reader): New subclass of rtx_reader.
(rtx_handle_directive): Convert to...
(gen_reader::handle_unknown_directive): ...this.
(init_rtx_reader_args_cb): Convert return type from bool to
rtx_reader *. Create a gen_reader instance, using it for the
call to read_md_files. Return it if no errors occur.
(init_rtx_reader_args): Convert return type from bool to
rtx_reader *.
* gensupport.h (init_rtx_reader_args_cb): Likewise.
(init_rtx_reader_args_cb): Likewise.
* read-md.c (struct file_name_list): Move to class rtx_reader.
(read_md_file): Delete in favor of rtx_reader::m_read_md_file.
(read_md_filename): Delete in favor of
rtx_reader::m_read_md_filename.
(read_md_lineno): Delete in favor of rtx_reader::m_read_md_lineno.
(in_fname): Delete in favor of rtx_reader::m_toplevel_fname.
(base_dir): Delete in favor of rtx_reader::m_base_dir.
(first_dir_md_include): Delete in favor of
rtx_reader::m_first_dir_md_include.
(last_dir_md_include_ptr): Delete in favor of
rtx_reader::m_last_dir_md_include_ptr.
(max_include_len): Delete.
(rtx_reader_ptr): New.
(fatal_with_file_and_line): Use get_filename and get_lineno
accessors of rtx_reader_ptr.
(require_char_ws): Likewise.
(rtx_reader::read_char): New method, based on ::read_char.
(rtx_reader::unread_char): New method, based on ::unread_char.
(read_escape): Use get_filename and get_lineno accessors of
rtx_reader_ptr.
(read_braced_string): Use get_lineno accessor of rtx_reader_ptr.
(read_string): Use get_filename and get_lineno accessors of
rtx_reader_ptr.
(rtx_reader::rtx_reader): New ctor.
(rtx_reader::~rtx_reader): New dtor.
(handle_include): Convert from a function to...
(rtx_reader::handle_include): ...this method, converting
handle_directive from a callback to a virtual function.
(handle_file): Likewise, converting to...
(rtx_reader::handle_file): ...this method.
(handle_toplevel_file): Likewise, converting to...
(rtx_reader::handle_toplevel_file): ...this method.
(rtx_reader::get_current_location): New method.
(parse_include): Convert from a function to...
(rtx_reader::add_include_path): ...this method, dropping redundant
update to unused max_include_len.
(read_md_files): Convert from a function to...
(rtx_reader::read_md_files): ...this method, converting
handle_directive from a callback to a virtual function.
(noop_reader::handle_unknown_directive): New method.
* read-md.h (directive_handler_t): Delete this typedef.
(in_fname): Delete.
(read_md_file): Delete.
(read_md_lineno): Delete.
(read_md_filename): Delete.
(class rtx_reader): New class.
(rtx_reader_ptr): New decl.
(class noop_reader): New subclass of rtx_reader.
(read_char): Reimplement in terms of rtx_reader::read_char.
(unread_char): Reimplement in terms of rtx_reader::unread_char.
(read_md_files): Delete.
* read-rtl.c (read_rtx_code): Update for deletion of globals
read_md_filename and read_md_lineno.
* c-common.c (c_common_truthvalue_conversion): Inhibit
Wint-in-bool-context warning with from_macro_definition_at.
Mention the expression will always evaluate to true.
Joseph Myers [Wed, 21 Sep 2016 17:52:47 +0000 (18:52 +0100)]
Add _FloatN, _FloatNx tests for __builtin_fpclassify.
This patch adds tests for _FloatN and _FloatNx arguments to the
__builtin_fpclassify type-generic built-in function, omitted from the
original tests for type-generic functions on these types.
Tested for x86_64-pc-linux-gnu; all the supported new tests pass.
Eric Botcazou [Wed, 21 Sep 2016 12:53:20 +0000 (12:53 +0000)]
configure.ac: Do not create links, only substitute the filenames.
* configure.ac: Do not create links, only substitute the filenames.
* configure: Regenerate.
* Makefile.in: Assign the substitution results to variables.
(LIBGCC_LINKS): Define.
(enable-execute-stack.c): New rule.
(unwind.h): Likewise.
(md-unwind-support.h): Likewise.
(sfp-machine.h): Likewise.
(gthr-default.h): Likewise.
Add $(LIBGCC_LINKS) to the prerequisites of all object files and
unwind.h as prerequisite of install-unwind_h-forbuild.
re PR target/77326 ([avr] Invalid optimization omits comparison)
gcc/
PR target/77326
* config/avr/avr.c (avr_notice_update_cc) [CC_NONE]: If insn
touches some regs mentioned in cc_status, do CC_STATUS_INIT.
gcc/testsuite/
PR target/77326
* gcc.target/avr/torture/pr77326.c: New test.
Richard Biener [Wed, 21 Sep 2016 07:38:33 +0000 (07:38 +0000)]
re PR tree-optimization/77648 (Setting conversion to a integer to double to 0 3/4 through a loop)
2016-09-21 Richard Biener <rguenther@suse.de>
PR tree-optimization/77648
* tree-ssa-structalias.c (process_constraint): Handle all DEREF
with complex RHS.
(make_transitive_closure_constraints): Adjust comment.
(make_any_offset_constraints): New function.
(handle_rhs_call): Make sure to first expand a pointer to all
subfields before transitively closing it.
(handle_const_call): Likewise. Properly expand returned
pointers as well.
(handle_pure_call): Likewise.
* gcc.dg/torture/pr77648-1.c: New testcase.
* gcc.dg/torture/pr77648-2.c: Likewise.
Paul Thomas [Wed, 21 Sep 2016 06:57:28 +0000 (06:57 +0000)]
re PR fortran/77657 (link error with implementation of user-defined derived type input/output (UD-DTIO) in child extending abstract parent)
2016-09-21 Paul Thomas <pault@gcc.gnu.org>
PR fortran/77657
* interface.c (gfc_find_specific_dtio_proc): Borrow trick from
resolve_typebound_generic_call to find dtio procedures that
over-ride those in the declared type.
2016-09-21 Paul Thomas <pault@gcc.gnu.org>
PR fortran/77657
* gfortran.dg/dtio_12.f90: New test.
For the lower vrp bound to be 2/-2, unsigned ints must be 4 bytes wide. This
breaks targets like avr. Explicitly using __UINT32_TYPE__ (behind a typedef)
makes the testcase pass for all targets.
PR tree-optimization/72835
* tree-ssa-reassoc.c (make_new_ssa_for_def): New.
(make_new_ssa_for_all_defs): Likewise.
(zero_one_operation): Replace all SSA_NAMEs defined in the chain.
Martin Sebor [Wed, 21 Sep 2016 01:39:27 +0000 (01:39 +0000)]
PR middle-end/49905 - Better sanity checking on sprintf src & dest to
gcc/ChangeLog:
PR middle-end/49905
* Makefile.in (OBJS): Add gimple-ssa-sprintf.o.
* config/linux.h (TARGET_PRINTF_POINTER_FORMAT): Redefine.
* config/linux.c (gnu_libc_printf_pointer_format): New function.
* config/sol2.h (TARGET_PRINTF_POINTER_FORMAT): Same.
* config/sol2.c (solaris_printf_pointer_format): New function.
* doc/invoke.texi (-Wformat-length, -fprintf-return-value): New
options.
* doc/tm.texi.in (TARGET_PRINTF_POINTER_FORMAT): Document.
* doc/tm.texi: Regenerate.
* gimple-fold.h (get_range_strlen): New function.
(get_maxval_strlen): Declare existing function.
* gimple-fold.c (get_range_strlen): Add arguments and compute both
maximum and minimum.
(get_range_strlen): Define overload.
(get_maxval_strlen): Adjust.
* gimple-ssa-sprintf.c: New file and pass.
* passes.def (pass_sprintf_length): Add new pass.
* targhooks.h (default_printf_pointer_format): Declare new function.
(gnu_libc_printf_pointer_format): Same.
(solaris_libc_printf_pointer_format): Same.
* targhooks.c (default_printf_pointer_format): Define new function.
* tree-pass.h (make_pass_sprintf_length): Declare new function.
* print-tree.c: Increase buffer size.
gcc/c-family/ChangeLog:
PR middle-end/49905
* c.opt: Add -Wformat-length and -fprintf-return-value.
gcc/testsuite/ChangeLog:
PR middle-end/49905
* gcc.dg/builtin-stringop-chk-1.c: Adjust.
* gcc.dg/tree-ssa/builtin-sprintf-warn-1.c: New test.
* gcc.dg/tree-ssa/builtin-sprintf-warn-2.c: New test.
* gcc.dg/tree-ssa/builtin-sprintf-warn-3.c: New test.
* gcc.dg/tree-ssa/builtin-sprintf-warn-4.c: New test.
* gcc.dg/tree-ssa/builtin-sprintf.c: New test.
* gcc.dg/tree-ssa/builtin-sprintf-2.c: New test.
* g++.dg/ipa/pure-const-3.C: Add -fno-ipa-vrp. Else constant arguments
will be optimized away.
* gcc.dg/ipa/vrp1.c: New test.
* gcc.dg/ipa/vrp2.c: New test.
* gcc.dg/ipa/vrp3.c: New test.
* g++.dg/tree-ssa/pr31146-2.C: Run with -fno-tree-evrp as evrp also
does the same transformation.
* g++.dg/warn/pr33738.C: XFAIL as optimization now happens in ccp.
* gcc.dg/tree-ssa/evrp1.c: New test.
* gcc.dg/tree-ssa/evrp2.c: New test.
* gcc.dg/tree-ssa/evrp3.c: New test.
* gcc.dg/tree-ssa/pr20657.c: Check for the pattern in evrp dump.
* gcc.dg/tree-ssa/pr22117.c: Likewise.
* gcc.dg/tree-ssa/pr61839_2.c: Likewise.
* gcc.dg/tree-ssa/pr64130.c: Likewise.
* gcc.dg/tree-ssa/pr37508.c: Change the pattern to be checked as
foling now happens early.
* gcc.dg/tree-ssa/vrp04.c: Likewise.
* gcc.dg/tree-ssa/vrp06.c: Likewise.
* gcc.dg/tree-ssa/vrp16.c: Likewise.
* gcc.dg/tree-ssa/vrp25.c: Likewise.
* gcc.dg/tree-ssa/vrp67.c: Likewise.
Ian Lance Taylor [Tue, 20 Sep 2016 17:52:23 +0000 (17:52 +0000)]
re PR go/77625 (go/gofrontend/ast-dump.cc:169:42: error: ‘new’ of type ‘std::ofstr eam {aka std::basic_ofstream<char>}’ with extended alignment 16)
PR go/77625
compiler: allocate std::ofstream as a local variable
GCC PR 77625 points out a warning about new std::ofstream. I don't know
how that is supposed to work, but in this case the std::ofstream may as
well be a local variable anyhow.
Jakub Jelinek [Tue, 20 Sep 2016 15:17:41 +0000 (17:17 +0200)]
re PR c++/77638 (ICE on x86_64-linux-gnu (internal compiler error: tree check: expected tree that contains ‘decl common’ structure, have ‘error_mark’ in cp_parser_template_declaration_after_parameters, at cp/parser.c:25722))
PR c++/77638
* parser.c (cp_parser_template_declaration_after_parameter): For 2
argument operator"" template set ok to false for
parm == error_mark_node.
re PR libstdc++/77619 (uninitialized_meow_construct and friends not exception safe)
PR libstdc++/77619
* include/bits/stl_construct.h: (_Construct_novalue): New.
(_Destroy_n_aux, _Destroy_n): New.
* include/bits/stl_uninitialized.h: (type_traits):
New include in C++11 mode.
(__uninitialized_default_novalue_1): New.
(__uninitialized_default_novalue_n_1): Likewise.
(__uninitialized_default_novalue): Likewise.
(__uninitialized_default_novalue_n): Likewise.
(__uninitialized_copy_n_pair): Likewise.
(uninitialized_default_construct):
Use __uninitialized_default_novalue.
(uninitialized_default_construct_n):
Use __uninitialized_default_novalue_n.
(uninitialized_value_construct): Use __uninitialized_default.
(uninitialized_value_construct_n): Use __uninitialized_default_n.
(uninitialized_move): Use uninitialized_copy.
(uninitialized_move_n): Use __uninitialized_copy_n_pair.
(destroy_at): Use _Destroy.
(destroy): Likewise.
(destroy_n): Likewise.
* testsuite/20_util/specialized_algorithms/
memory_management_tools/1.cc: Add tests for exceptions,
add tests for trivial cases for construct and move.
Joseph Myers [Mon, 19 Sep 2016 21:53:30 +0000 (22:53 +0100)]
Define TS 18661-1 CR_DECIMAL_DIG in <float.h>.
TS 18661-1 defines a macro CR_DECIMAL_DIG in <float.h>, for the number
of decimal digits for which conversions between decimal character
strings and (IEEE) binary formats, in both directions, are correctly
rounded. This patch implements support for this macro in GCC's
<float.h>.
The definition __UINTMAX_MAX__ is the right one for GCC's conversions
of floating constants, since I made those use MPFR to make them
correctly rounding. The macro also covers standard library functions
such as strtod and printf. The definition is also correct for current
glibc. If any targets' libcs support correct rounding in a way that
conforms to TS 18661-1 with a smaller value of CR_DECIMAL_DIG, making
<float.h> reflect that could not be done in isolation without changes
to the interpretation of floating constants as well, since a smaller
CR_DECIMAL_DIG requires double rounding of floating constants (first
to CR_DECIMAL_DIG decimal digits, then to the desired binary format).
Boostrapped with no regressions on x86_64-pc-linux-gnu.
gcc:
* ginclude/float.h [__STDC_WANT_IEC_60559_BFP_EXT__]
(CR_DECIMAL_DIG): New macro.
gcc/testsuite:
* gcc.dg/cr-decimal-dig-1.c: New test.
Joseph Myers [Mon, 19 Sep 2016 21:51:56 +0000 (22:51 +0100)]
Make max_align_t respect _Float128.
The _FloatN, _FloatNx, _DecimalN and _DecimalNx types are specified in
such a way that they are basic types, meaning that max_align_t must be
at least as aligned as those types.
On 32-bit x86, max_align_t is currently 8-byte aligned, but
_Decimal128 and _Float128 are 16-byte aligned, so the alignment of
max_align_t needs to increase to meet the standard requirements for
these types.
This patch implements such an increase. Because max_align_t needs to
be usable for C++ as well as for C, <stddef.h> can't actually refer to
_Float128, but needs to use __float128 (or some other notation for the
type) instead. And since __float128 is architecture-specific, there
isn't a preprocessor conditional that means "__float128 is available"
(whereas one could test __FLT128_MANT_DIG__ to see if _Float128 is
available; __SIZEOF_FLOAT128__ is available on x86 only). But I
believe the only case that actually has an alignment problem here is
32-bit x86, and <stddef.h> already has lots of conditional specific to
particular architectures or OSes, so this patch uses a conditional on
__i386__; that also is the minimal change that ensures neither size
nor alignment of max_align_t is changed in any case other than where
it is necessary. If any other architectures turn out to have such an
issue, it will show up as failures of one of the testcases added by
this patch.
Such an increase is of course an ABI change, but a reasonably safe
one, in that max_align_t doesn't tend to appear in library interfaces
(rather, it's something to use when writing allocators and similar
code; most matches found on codesearch.debian.net look like copies of
the gnulib stddef.h module rather than anything actually using
max_align_t at all).
max_align_t_align has a corresponding change (adding _Float128 to the
types considered).
(I think glibc malloc alignment should also increase to 16-byte on
32-bit x86 so it works for allocating objects of these types, which is
now straightforward given the fix made for 32-bit powerpc.)
Bootstrapped with no regressions on x86_64-pc-linux-gnu, and
spot-tested with -m32 that the new float128-align.c test now compiles
where previously it didn't.