A performance regression was introduced by commit
84d74e427a "powerpc: Cleanup fenv_private.h".
In the powerpc implementation of SET_RESTORE_ROUND, there is the
following code in the "SET" function (slightly simplified):
--
old.fenv = fegetenv_register ();
new.l = (old.l & _FPU_MASK_TRAPS_RN) | r; (1)
if (new.l != old.l) (2)
{
if ((old.l & _FPU_ALL_TRAPS) != 0)
(void) __fe_mask_env ();
fesetenv_register (new.fenv); (3)
--
Line (1) sets the value of "new" to the current value of FPSCR,
but masks off summary bits, exceptions, non-IEEE mode, and
rounding mode, then ORs in the new rounding mode.
Line (2) compares this new value to the current value in order to
avoid setting a new value in the FPSCR (line (3)) unless something
significant has changed (exception enables or rounding mode).
The summary bits are not germane to the comparison, but are cleared
in "new" and preserved in "old", resulting in false negative
comparisons, and unnecessarily setting the FPSCR in those cases
with associated negative performance impacts.
The solution is to treat the summaries identically for "new" and "old":
- save them in SET
- leave them alone otherwise
- restore the saved values in RESTORE
Also minor changes:
- expand _FPU_MASK_RN to 64bit hex, to match other MASKs
- treat bit 52 (left-to-right) as reserved (since it is)
* sysdeps/powerpc/fpu/fenv_private.h (_FPU_MASK_TRAPS_RN):
(_FPU_MASK_FRAC_INEX_RET_CC): Fix masks to more properly handle
summary bits.
(_FPU_MASK_RN): Expand _FPU_MASK_RN to 64bit hex.
(_FPU_MASK_NOT_RN_NI): Treat bit 52 (left-to-right) as reserved.
Reviewed-by: Tulio Magno Quites Machado Filho <tuliom@linux.vnet.ibm.com>
This patch moves the generic definition from x86_64 init-arch
to a common header ifunc-init.h. No functional changes is expected.
Checked on a x86_64-linux-gnu build.
* sysdeps/generic/ifunc-init.h: New file.
* sysdeps/x86/init-arch.h: Use generic ifunc-init.h.
With support for _Float128 functions on platforms where that type has
the same ABI as long double, as well as on platforms where it is
ABI-distinct, those functions will need to be exported from glibc's
shared libraries at appropriate symbol versions in each case.
This patch avoids duplication of lists of symbols to export by moving
the symbols other than __* to math/Versions and stdlib/Versions.
There, they are conditional on <float128-abi.h> defining
FLOAT128_VERSION and a default version of that header is added that
does not define that macro. Enabling the float128 function aliases
will then include adding a sysdeps/ieee754/ldbl-128/float128-abi.h
that defines FLOAT128_VERSION to GLIBC_2.27. Symbols __* remain in
sysdeps/ieee754/float128/Versions; those symbols should be present
only once per floating-point format, not once per type.
Note that if any platforms currently lacking support for a type with
binary128 format get glibc support for such a type in future (whether
only as _Float128, or also as a new long double format), and new libm
functions (present for all types) have been added by then, additional
macros will be needed to allow such functions to get a version of the
form "GLIBC_2.28 if the platform had _Float128 support by then, or the
later version at which that platform had _Float128 support added".
This is not however a preexisting condition, but would have applied
equally to the existing support for _Float128 as an ABI-distinct
type. New all-type libm functions should just be added to the
appropriate symbol version (currently GLIBC_2.27) for all types, with
such special-case handling for _Float128 versions (and _Float64x as
well in future) waiting until someone actually wants to add support
for _Float128 to an existing platform after a release in which that
platform and a post-2.26 libm function had support but that platform
lacked _Float128 support.
Tested with build-many-glibcs.py that installed stripped shared
libraries are unchanged by this patch. Also tested in conjunction
with the remaining changes to enable float128 aliases.
* sysdeps/generic/float128-abi.h: New file.
* sysdeps/ieee754/float128/Versions (FLOAT128_VERSION): Move
non-__prefixed symbols to ....
* math/Versions: ... here. Include <float128-abi.h>.
* stdlib/Versions ... and here. Include <float128-abi.h>
This patch adds support for building strtof128, wcstof128, strtof128_l
and wcstof128_l as aliases, in the case of __HAVE_FLOAT128 &&
!__HAVE_DISTINCT_FLOAT128.
Tested with build-many-glibcs.py that installed stripped shared
libraries are unchanged by this patch. Also tested together with
changes to enable float128 aliases.
* stdlib/strtold.c: Include <bits/floatn.h>
[__HAVE_FLOAT128 && !__HAVE_DISTINCT_FLOAT128] (strtof128): Define
and later undefine as macro. Define as weak alias if
[!USE_WIDE_CHAR].
[__HAVE_FLOAT128 && !__HAVE_DISTINCT_FLOAT128] (wcstof128): Define
and later undefine as macro. Define as weak alias if
[USE_WIDE_CHAR].
* sysdeps/ieee754/ldbl-128/strtold_l.c [__HAVE_FLOAT128 &&
!__HAVE_DISTINCT_FLOAT128] (strtof128_l): Define and later
undefine as macro. Define as weak alias if [!USE_WIDE_CHAR].
[__HAVE_FLOAT128 && !__HAVE_DISTINCT_FLOAT128] (wcstof128_l):
Define and later undefine as macro. Define as weak alias if
[USE_WIDE_CHAR].
* sysdeps/ieee754/ldbl-64-128/strtold_l.c: Include
<bits/floatn.h>.
[__HAVE_FLOAT128 && !__HAVE_DISTINCT_FLOAT128] (strtof128_l):
Define and later undefine as macro. Define as weak alias if
[!USE_WIDE_CHAR].
[__HAVE_FLOAT128 && !__HAVE_DISTINCT_FLOAT128] (wcstof128_l):
Define and later undefine as macro. Define as weak alias if
[USE_WIDE_CHAR].
This patch makes ldbl-64-128/s_nextafterl.c restore the default
weak_alias definition and use libm_alias_ldouble_other (having
undefined and redefined weak_alias for the include of
ldbl-128/s_nextafterl.c, so the libm_alias_ldouble use in the latter
file is ineffective).
Tested with build-many-glibcs.py that installed stripped shared
libraries are unchanged by this patch. Also tested together with
changes to enable float128 aliases.
* sysdeps/ieee754/ldbl-64-128/s_nextafterl.c (weak_alias):
Undefine and restore default definition. Use
libm_alias_ldouble_other.
Normally, TLS relocations against local symbols are optimised by the linker
to be absolute. However, gold does not do this, and so it is possible to
end up with, for example, R_SPARC_TLS_DTPMOD64 referring to a local symbol.
Since sym_map is left as null in elf_machine_rela for the special local
symbol case, the relocation handling thinks it has nothing to do, and so
the module gets left as 0. Havoc then ensues when the variable in question
is accessed.
Before this fix, the main_local_gold program would receive a SIGBUS on
sparc64, and SIGSEGV on powerpc32. With this fix applied, that test now
passes like the rest of them.
* sysdeps/powerpc/powerpc32/dl-machine.h (elf_machine_rela):
Assign sym_map to be map for local symbols, as TLS relocations
use sym_map to determine whether the symbol is defined and to
extract the TLS information.
* sysdeps/sparc/sparc32/dl-machine.h (elf_machine_rela): Likewise.
* sysdeps/sparc/sparc64/dl-machine.h (elf_machine_rela): Likewise.
Fix the ifdef clause that was being used in the opposite way, setting
a wrong value of the carry bit.
This is also correcting 2 memory accesses that were mistakenly referring
to r0 while they were supposed to mean the immediate value 0.
[BZ #22142]
* stdio-common/tst-printf.c (fp_test): Add tests for DBL_MAX and
-DBL_MAX.
(do_test): Likewise.
* stdio-common/tst-printf.sh: Likewise.
* sysdeps/powerpc/powerpc64/power7/add_n.S: Invert the initial
ifdef clause in order to set the carry bit right. Replace r0 by
0 without changing the behavior.
This patch makes SPARC fabsl implementation use libm_alias_ldouble, to
prepare them for also defining _Float128 function aliases.
Tested with build-many-glibcs.py that installed stripped shared
libraries (sparc64-linux-gnu and sparcv9-linux-gnu) are unchanged by
the patch.
* sysdeps/sparc/sparc32/fpu/s_fabsl.c: Include
<libm-alias-ldouble.h>.
(fabsl): Define using libm_alias_ldouble.
* sysdeps/sparc/sparc64/fpu/s_fabsl.c: Include
<libm-alias-ldouble.h>.
(fabsl): Define using libm_alias_ldouble.
Testing with changes to enable _Float128 function aliases shows that
the libm_alias_ldouble_other usage in ldbl-opt/w_lgamma_compatl.c does
not in fact work. Furthermore, it is unnecessary; the relevant
aliases get created through w_lgammal_compat2.c. This patch removes
the problem code.
Tested with build-many-glibcs.py that installed stripped shared
libraries are unchanged by the patch. Also tested in conjunction with
patches to enable _Float128 function aliases.
* sysdeps/ieee754/ldbl-opt/w_lgamma_compatl.c [BUILD_LGAMMA]:
Remove conditional code.
Testing with changes to enable _Float128 function aliases shows that
the libm_alias_ldouble_other usage in ldbl-opt/s_clog10l.c does not in
fact work, because __clog10l is defined with long_double_symbol rather
than as a normal C alias. This patch fixes this by renaming the
__clog10l__internal alias (not strictly necessary, but avoids a hack
with "__clog10l_interna" / "__clog10l__interna" as first argument to
libm_alias_ldouble_other) and using the renamed alias when calling
libm_alias_ldouble_other.
Tested with build-many-glibcs.py that installed stripped shared
libraries are unchanges by the patch. Also tested in conjunction with
patches to enable _Float128 function aliases.
* sysdeps/ieee754/ldbl-opt/s_clog10l.c (__clog10l__internal):
Rename to __clog10_internal_l.
(__clog10_internal_l): Define aliases using
libm_alias_ldouble_other instead of using libm_alias_ldouble_other
with __clog10.
Current GLIBC has two ways to implement the single thread optimization
on syscalls to avoid calling the cancellation path: either by using
global variables (__{libc,pthread}_multiple_thread) or by accessing
the TCB field (defined by TLS_MULTIPLE_THREADS_IN_TCB). Both the
variables and the macros to acces its value are defined in the
architecture sysdep-cancel.h header.
This patch consolidates its definition on only one header,
sysdeps/unix/sysv/linux/sysdep-cancel.h, and adds a new define
(SINGLE_THREAD_BY_GLOBAL) which the architecture defines if it prefer
to use the global variables instead of the TCB field. This is an
optimization, so if the architecture does not define it, the TCB
method will be used as default.
Checked on x86_64-linux-gnu and on a build with major touched
ABIs (aarch64-linux-gnu, alpha-linux-gnu, arm-linux-gnueabihf,
hppa-linux-gnu, i686-linux-gnu, m68k-linux-gnu, microblaze-linux-gnu,
mips-linux-gnu, mips64-linux-gnu, powerpc-linux-gnu,
powerpc64le-linux-gnu, s390-linux-gnu, s390x-linux-gnu, sh4-linux-gnu,
sparcv9-linux-gnu, sparc64-linux-gnu, tilegx-linux-gnu).
* sysdeps/unix/sysv/linux/aarch64/sysdep-cancel.h: Remove file.
* sysdeps/unix/sysv/linux/alpha/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/arm/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/hppa/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/mips/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/nios2/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/powerpc/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/s390/s390-32/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/s390/s390-64/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/sh/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/sparc/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/tile/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/x86_64/sysdep-cancel.h: Likewise.
* sysdeps/unix/sysv/linux/s390/s390-64/sysdep.h
(SINGLE_THREAD_BY_GLOBAL): Define.
* sysdeps/unix/sysv/linux/aarch64/sysdep.h (SINGLE_THREAD_BY_GLOBAL):
Likewise.
* sysdeps/unix/sysv/linux/alpha/sysdep.h (SINGLE_THREAD_BY_GLOBAL):
Likewise.
* sysdeps/unix/sysv/linux/arm/sysdep.h (SINGLE_THREAD_BY_GLOBAL):
Likewise.
* sysdeps/unix/sysv/linux/hppa/sysdep.h (SINGLE_THREAD_BY_GLOBAL):
Likewise.
* sysdeps/unix/sysv/linux/microblaze/sysdep.h
(SINGLE_THREAD_BY_GLOBAL): Likewise.
* sysdeps/unix/sysv/linux/x86_64/sysdep.h (SINGLE_THREAD_BY_GLOBAL):
Likewise.
This patch fixes ldbl-opt code to use generic libm alias macros in
preparation for getting _FloatN / _FloatNx aliases where appropriate.
Four functions are affected, that undefine and redefine alias macros
before including the implementations they wrap in such a way that
_FloatN / _FloatNx aliases would not appear. s_clog10l.c undefines
and redefined declare_mgen_alias, so just needs a
libm_alias_ldouble_other call added. w_exp10l_compat.c undefines and
redefines weak_alias, but in fact does not need to do so, since
math/w_exp10l_compat.c uses libm_alias_ldouble and does not use
weak_alias other than through that, so the undefines and redefines of
weak_alias are removed. w_lgamma_compatl.c and w_remainderl_compat.c
are made to use libm_alias_ldouble_other in conjunction with restoring
the original definition of weak_alias so this is effective.
Tested with build-many-glibcs.py. Installed stripped shared libraries
are unchanged by this patch.
* sysdeps/ieee754/ldbl-opt/s_clog10l.c: Use
libm_alias_ldouble_other.
* sysdeps/ieee754/ldbl-opt/w_exp10l_compat.c (weak_alias): Do not
undefine and redefine.
[LIBM_SVID_COMPAT && !LONG_DOUBLE_COMPAT (libm, GLIBC_2_1)]
(exp10l): Do not define here.
* sysdeps/ieee754/ldbl-opt/w_lgamma_compatl.c [BUILD_LGAMMA]
(weak_alias): Undefine and redefine.
[BUILD_LGAMMA]: Use libm_alias_ldouble_other.
* sysdeps/ieee754/ldbl-opt/w_remainderl_compat.c
[LIBM_SVID_COMPAT] (weak_alias): Undefine and redefine here.
[LIBM_SVID_COMPAT]: Use libm_alias_ldouble_other.
Some libm functions are unable to use the generic alias macros such as
libm_alias_double because they have special symbol versioning
requirements for the main float, double or long double public names.
To facilitate adding _FloatN / _FloatNx function aliases in future,
it's still desirable to have generic macros those functions can use as
far as possible. This patch adds macros such as
libm_alias_double_other, which only define names for _FloatN /
_FloatNx aliases, not for float / double / long double. As present,
all these new macros do nothing, but they are called in the
appropriate places in macros such as libm_alias_double. This patch
also arranges for lgamma implementations, and the recently added
optimized float function implementations, to use the new macros to
make them ready for addition of _FloatN / _FloatNx aliases.
Tested for x86_64, and tested with build-many-glibcs.py that installed
stripped shared libraries are unchanged by this patch.
* sysdeps/generic/libm-alias-double.h (libm_alias_double_other_r):
New macro.
(libm_alias_double_other): Likewise.
(libm_alias_double_r): Use libm_alias_double_other_r.
* sysdeps/generic/libm-alias-float.h (libm_alias_float_other_r):
New macro.
(libm_alias_float_other): Likewise.
(libm_alias_float_r): Use libm_alias_float_other_r.
* sysdeps/generic/libm-alias-float128.h
(libm_alias_float128_other_r): New macro.
(libm_alias_float128_other): Likewise.
(libm_alias_float128_r): Use libm_alias_float128_other_r.
* sysdeps/generic/libm-alias-ldouble.h
(libm_alias_ldouble_other_r): New macro.
(libm_alias_ldouble_other): Likewise.
(libm_alias_ldouble_r): Use libm_alias_ldouble_other_r.
* sysdeps/ieee754/ldbl-opt/libm-alias-double.h
(libm_alias_double_other_r): New macro.
(libm_alias_double_other): Likewise.
(libm_alias_double_r): Use libm_alias_double_other_r.
* sysdeps/ieee754/ldbl-opt/libm-alias-ldouble.h
(libm_alias_ldouble_other_r): New macro.
(libm_alias_ldouble_other): Likewise.
(libm_alias_ldouble_r): Use libm_alias_ldouble_other_r.
* math/w_lgamma_main.c: Include <libm-alias-double.h>.
[!USE_AS_COMPAT]: Use libm_alias_double_other.
* math/w_lgammaf_main.c: Include <libm-alias-float.h>.
[!USE_AS_COMPAT]: Use libm_alias_float_other.
* math/w_lgammal_main.c: Include <libm-alias-ldouble.h>.
[!USE_AS_COMPAT]: Use libm_alias_ldouble_other.
* math/w_exp2f.c: Use libm_alias_float_other.
* math/w_expf.c: Likewise.
* math/w_log2f.c: Likewise.
* math/w_logf.c: Likewise.
* math/w_powf.c: Likewise.
* sysdeps/ieee754/flt-32/e_exp2f.c: Include <libm-alias-float.h>.
[!__exp2f]: Use libm_alias_float_other.
* sysdeps/ieee754/flt-32/e_expf.c: Include <libm-alias-float.h>.
[!__expf]: Use libm_alias_float_other.
* sysdeps/ieee754/flt-32/e_log2f.c: Include <libm-alias-float.h>.
[!__log2f]: Use libm_alias_float_other.
* sysdeps/ieee754/flt-32/e_logf.c: Include <libm-alias-float.h>.
[!__logf]: Use libm_alias_float_other.
* sysdeps/ieee754/flt-32/e_powf.c: Include <libm-alias-float.h>.
[!__powf]: Use libm_alias_float_other.
Continuing the use of generic macros for defining libm function
aliases, in preparation for adding more _FloatN / _FloatNx function
names, this patch makes the lgamma_r functions use such macros.
declare_mgen_alias_r becomes a standard macro in math-type-macros.h
instead of being locally defined in w_lgamma_r_templace.c. This in
turn must be defined by each math-type-macros-<type>.h. Rather than
providing an unused default in math-type-macros.h, that header is made
to give an error if math-type-macros-<type>.h failed to define
declare_mgen_alias or declare_mgen_alias_r. The compat lgamma_r
wrappers are updated similarly. The ldbl-opt versions are removed as
no longer needed.
Tested for x86_64, and with build-many-glibcs.py. Installed stripped
shared libraries are unchanged except for powerpc64le (where the usual
issue applies that an ldbl-opt long double function previously used
long_double_symbol unconditionally and now the symbol versions on
powerpc64le mean weak_alias is used instead, resulting in the same
symbol versions in the final shared library but still enough
difference in the input objects for that library not to be
byte-identical).
* sysdeps/generic/math-type-macros.h [!declare_mgen_alias]: Give
error. Remove default definition of declare_mgen_alias.
[!declare_mgen_alias_r]: Likewise.
* sysdeps/generic/math-type-macros-double.h
[!declare_mgen_alias_r] (declare_mgen_alias_r): New macro.
* sysdeps/generic/math-type-macros-float.h [!declare_mgen_alias_r]
(declare_mgen_alias_r): Likewise.
* sysdeps/generic/math-type-macros-float128.h
[!declare_mgen_alias_r] (declare_mgen_alias_r): Likewise.
* sysdeps/generic/math-type-macros-ldouble.h
[!declare_mgen_alias_r] (declare_mgen_alias_r): Likewise.
* math/w_lgamma_r_template.c (declare_mgen_alias_r_x): Remove
macro.
(declare_mgen_alias_r_s): Likewise.
(declare_mgen_alias_r): Likewise.
* math/w_lgamma_r_compat.c: Include <libm-alias-double.h>.
(lgamma_r): Define using libm_alias_double_r.
* math/w_lgammaf_r_compat.c: Include <libm-alias-float.h>.
(lgammaf_r): Define using libm_alias_float_r.
* math/w_lgammal_r_compat.c: Include <libm-alias-ldouble.h>.
(lgammal_r): Define using libm_alias_ldouble_r.
* sysdeps/ieee754/ldbl-opt/w_lgamma_r_compat.c: Remove file.
* sysdeps/ieee754/ldbl-opt/w_lgammal_r_compat.c: Likewise.
The ldbl-opt version of w_scalbln.c is not in fact needed; it handles
compat symbol versions for libc, but this file isn't built for libc,
only for libm. This patch removes this file.
Tested with build-many-glibcs.py that installed stripped shared
libraries are unchanged by this patch.
* sysdeps/ieee754/ldbl-opt/w_scalbln.c: Remove file.
This patch makes the ldbl-128 and ldbl-96 implementations of fma use
libm_alias_double.
Tested for x86_64, and tested with build-many-glibcs.py that installed
stripped shared libraries are unchanged by the patch.
* sysdeps/ieee754/ldbl-128/s_fma.c: Include <libm-alias-double.h>.
[!__fma] (fma): Define using libm_alias_double.
* sysdeps/ieee754/ldbl-96/s_fma.c: Include <libm-alias-double.h>.
[!__fma] (fma): Define using libm_alias_double.
This patch makes ldbl-128 functions use libm_alias_ldouble to define
function aliases. float128_private.h is updated accordingly. Most of
the ldbl-64-128 wrappers are removed as no longer needed with this
change (leaving those that involve versioning for functions in libc or
that shouldn't be exported from libm for _Float128 / _Float64x types
with the same format as long double).
Tested for x86_64, and tested with build-many-glibcs.py that installed
stripped shared libraries are unchanged by this patch.
* sysdeps/ieee754/float128/float128_private.h: Include
<libm-alias-ldouble.h> and <libm-alias-float128.h>.
(libm_alias_ldouble_r): Undefine and redefine.
* sysdeps/ieee754/ldbl-128/s_asinhl.c: Include
<libm-alias-ldouble.h>.
(asinhl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_atanl.c: Include
<libm-alias-ldouble.h>.
(atanl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_cbrtl.c: Include
<libm-alias-ldouble.h>.
(cbrtl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_ceill.c: Include
<libm-alias-ldouble.h>.
(ceill): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_copysignl.c: Include
<libm-alias-ldouble.h>.
(copysignl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_cosl.c: Include
<libm-alias-ldouble.h>.
(cosl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_erfl.c: Include
<libm-alias-ldouble.h>.
(erfl): Define using libm_alias_ldouble.
(erfcl): Likewise.
* sysdeps/ieee754/ldbl-128/s_expm1l.c: Include
<libm-alias-ldouble.h>.
(expm1l): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_fabsl.c: Include
<libm-alias-ldouble.h>.
(fabsl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_floorl.c: Include
<libm-alias-ldouble.h>.
(floorl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_fmal.c: Include
<libm-alias-ldouble.h>.
(fmal): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_frexpl.c: Include
<libm-alias-ldouble.h>.
(frexpl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_fromfpl.c (fromfpl): Define using
libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_fromfpl_main.c: Include
<libm-alias-ldouble.h>.
* sysdeps/ieee754/ldbl-128/s_fromfpxl.c (fromfpxl): Define using
libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_getpayloadl.c: Include
<libm-alias-ldouble.h>.
(getpayloadl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_llrintl.c: Include
<libm-alias-ldouble.h>.
(llrintl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_llroundl.c: Include
<libm-alias-ldouble.h>.
(llroundl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_logbl.c: Include
<libm-alias-ldouble.h>.
(logbl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_lrintl.c: Include
<libm-alias-ldouble.h>.
(lrintl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_lroundl.c: Include
<libm-alias-ldouble.h>.
(lroundl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_modfl.c: Include
<libm-alias-ldouble.h>.
(modfl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_nearbyintl.c: Include
<libm-alias-ldouble.h>.
(nearbyintl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_nextafterl.c: Include
<libm-alias-ldouble.h>.
(nextafterl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_nextupl.c: Include
<libm-alias-ldouble.h>.
(nextupl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_remquol.c: Include
<libm-alias-ldouble.h>.
(remquol): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_rintl.c: Include
<libm-alias-ldouble.h>.
(rintl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_roundevenl.c: Include
<libm-alias-ldouble.h>.
(roundevenl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_roundl.c: Include
<libm-alias-ldouble.h>.
(roundl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_setpayloadl.c (setpayloadl): Define
using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_setpayloadl_main.c: Include
<libm-alias-ldouble.h>.
* sysdeps/ieee754/ldbl-128/s_setpayloadsigl.c (setpayloadsigl):
Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_sincosl.c: Include
<libm-alias-ldouble.h>.
(sincosl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_sinl.c: Include
<libm-alias-ldouble.h>.
(sinl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_tanhl.c: Include
<libm-alias-ldouble.h>.
(tanhl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_tanl.c: Include
<libm-alias-ldouble.h>.
(tanl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_totalorderl.c: Include
<libm-alias-ldouble.h>.
(totalorderl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_totalordermagl.c: Include
<libm-alias-ldouble.h>.
(totalordermagl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_truncl.c: Include
<libm-alias-ldouble.h>.
(truncl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_ufromfpl.c (ufromfpl): Define using
libm_alias_ldouble.
* sysdeps/ieee754/ldbl-128/s_ufromfpxl.c (ufromfpxl): Define using
libm_alias_ldouble.
* sysdeps/ieee754/ldbl-64-128/s_copysignl.c: Include
<libm-alias-ldouble.h>.
(weak_alias): Do not undefine and redefine.
[IS_IN (libc)] (libm_alias_ldouble): Undefine and redefine.
(copysignl): Define with long_double_symbol only if [IS_IN
(libc)].
* sysdeps/ieee754/ldbl-64-128/s_frexpl.c: Include
<libm-alias-ldouble.h>.
(weak_alias): Do not undefine and redefine.
[IS_IN (libc)] (libm_alias_ldouble): Undefine and redefine.
(frexpl): Define with long_double_symbol only if [IS_IN (libc)].
* sysdeps/ieee754/ldbl-64-128/s_modfl.c: Include
<libm-alias-ldouble.h>.
(weak_alias): Do not undefine and redefine.
[IS_IN (libc)] (libm_alias_ldouble): Undefine and redefine.
(modfl): Define with long_double_symbol only if [IS_IN (libc)].
* sysdeps/ieee754/ldbl-64-128/s_asinhl.c: Remove file.
* sysdeps/ieee754/ldbl-64-128/s_atanl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_cbrtl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_ceill.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_cosl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_erfl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_expm1l.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_fabsl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_floorl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_fmal.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_llrintl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_llroundl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_logbl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_lrintl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_lroundl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_nearbyintl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_remquol.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_rintl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_roundl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_sincosl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_sinl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_tanhl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_tanl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_truncl.c: Likewise.
Various source files in ldbl-64-128 are redundant, because they wrap
files that no longer provide public symbols that need special
versioning (those symbols having moved to separate errno-setting
wrappers), or, in the case of w_scalblnl.c, because the type-generic
template now does everything required (it deals with symbol versioning
for use in libm, and this file is never built for libc anyway - the
compat scalbln* symbols in libc, as opposed to scalbn*, are only for
i386 and m68k and are aliases to the corresponding scalbn* symbols).
This patch removes those redundant files.
Tested with build-many-glibcs.py (for all ldbl-64-128 configurations)
that installed stripped shared libraries are unchanged by this patch.
* sysdeps/ieee754/ldbl-64-128/e_ilogbl.c: Remove file.
* sysdeps/ieee754/ldbl-64-128/s_log1pl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_scalblnl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/s_scalbnl.c: Likewise.
* sysdeps/ieee754/ldbl-64-128/w_scalblnl.c: Likewise.
Recent commit 59ba2d2b54 missed to add __memrchr_power8 in
ifunc list. Also handled discarding unwanted bytes for
unaligned inputs in power8 optimization.
2017-10-05 Rajalakshmi Srinivasaraghavan <raji@linux.vnet.ibm.com>
* sysdeps/powerpc/powerpc64/multiarch/memrchr-ppc64.c: Revert
back to powerpc32 file.
* sysdeps/powerpc/powerpc64/multiarch/memrchr.c
(memrchr): Add __memrchr_power8 to ifunc list.
* sysdeps/powerpc/powerpc64/power8/memrchr.S: Mask
extra bytes for unaligned inputs.
This patch makes ldbl-96 functions use libm_alias_ldouble to define
function aliases.
Tested for x86_64, and tested with build-many-glibcs.py that installed
stripped shared libraries are unchanged by the patch.
* sysdeps/ieee754/ldbl-96/s_asinhl.c: Include
<libm-alias-ldouble.h>.
(asinhl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_cbrtl.c: Include
<libm-alias-ldouble.h>.
(cbrtl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_copysignl.c: Include
<libm-alias-ldouble.h>.
(copysignl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_cosl.c: Include
<libm-alias-ldouble.h>.
(cosl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_erfl.c: Include
<libm-alias-ldouble.h>.
(erfl): Define using libm_alias_ldouble.
(erfcl): Likewise.
* sysdeps/ieee754/ldbl-96/s_fmal.c: Include
<libm-alias-ldouble.h>.
(fmal): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_frexpl.c: Include
<libm-alias-ldouble.h>.
(frexpl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_fromfpl.c (fromfpl): Define using
libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_fromfpl_main.c: Include
<libm-alias-ldouble.h>.
* sysdeps/ieee754/ldbl-96/s_fromfpxl.c (fromfpxl): Define using
libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_getpayloadl.c: Include
<libm-alias-ldouble.h>.
(getpayloadl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_llrintl.c: Include
<libm-alias-ldouble.h>.
(llrintl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_llroundl.c: Include
<libm-alias-ldouble.h>.
(llroundl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_lrintl.c: Include
<libm-alias-ldouble.h>.
(lrintl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_lroundl.c: Include
<libm-alias-ldouble.h>.
(lroundl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_modfl.c: Include
<libm-alias-ldouble.h>.
(modfl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_nextupl.c: Include
<libm-alias-ldouble.h>.
(nextupl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_remquol.c: Include
<libm-alias-ldouble.h>.
(remquol): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_roundevenl.c: Include
<libm-alias-ldouble.h>.
(roundevenl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_roundl.c: Include
<libm-alias-ldouble.h>.
(roundl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_setpayloadl.c (setpayloadl): Define
using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_setpayloadl_main.c: Include
<libm-alias-ldouble.h>.
* sysdeps/ieee754/ldbl-96/s_setpayloadsigl.c: Include
<libm-alias-ldouble.h>.
(setpayloadsigl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_sincosl.c: Include
<libm-alias-ldouble.h>.
(sincosl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_sinl.c: Include
<libm-alias-ldouble.h>.
(sinl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_tanhl.c: Include
<libm-alias-ldouble.h>.
(tanhl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_tanl.c: Include
<libm-alias-ldouble.h>.
(tanl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_totalorderl.c: Include
<libm-alias-ldouble.h>.
(totalorderl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_totalordermagl.c: Include
<libm-alias-ldouble.h>.
(totalordermagl): Define using libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_ufromfpl.c (ufromfpl): Define using
libm_alias_ldouble.
* sysdeps/ieee754/ldbl-96/s_ufromfpxl.c (ufromfpxl): Define using
libm_alias_ldouble.
This is an optimized memmove implementation for the Qualcomm Falkor
processor core. Due to the way the falkor memcpy needs to be written,
code cannot be easily shared between memmove and memcpy like in case
of other aarch64 memcpy implementations due to which this routine is
separate. The underlying principle is the same as that of memcpy
where it tries to use registers with the same lower 4 bits for
fetching the same stream, thus optimizing hardware prefetcher
performance.
The memcpy copy loop copies 64 bytes at a time using the same register
pair since that's the way to train the hardware prefetcher on the
falkor core. memmove cannot quite do that since it needs to avoid
overlaps, so it does the next best thing, i.e. has a 32 byte loop with
a 32 byte end (prefetch a loop ahead to account for overlapping
locations) with register pairs that alias so that they hit the same
prefetcher. Due to this difference in loop size, they have to
currently be separate implementations but efforts are on to try and
get memmove to fall back into memcpy whenever it can without simply
duplicating all of the code.
Performance:
The routine fares around 20-25% better than the generic memmove for
most medium to large sizes (i.e. > 128 bytes) for the new walking
memmove benchmark (memmove-walk) with an unexplained regression
between 1K and 2K. The minor regression is something worth looking
into for us, but the remaining gains are significant enough that we
would like this included upstream as we looking into the cause for the
regression. Here is a snippet of the numbers as generated from the
microbenchmark by the compare_strings script. Comparisons are against
__memmove_generic:
Function: memmove
Variant: walk
__memmove_thunderx __memmove_falkor __memmove_generic
========================================================================================================================
<snip>
length=16384: 12508800.00 ( 6.09%) 11486800.00 ( 13.76%) 13319600.00
length=16400: 13614200.00 ( -0.67%) 11585000.00 ( 14.33%) 13523600.00
length=16385: 13448400.00 ( 0.10%) 11732700.00 ( 12.84%) 13461200.00
length=16399: 13594100.00 ( -0.22%) 11859600.00 ( 12.57%) 13564400.00
length=16386: 13211600.00 ( 1.13%) 11503800.00 ( 13.91%) 13362400.00
length=16398: 13218600.00 ( 2.12%) 11573200.00 ( 14.30%) 13504700.00
length=16387: 13510900.00 ( -0.37%) 11744200.00 ( 12.76%) 13461300.00
length=16397: 13603700.00 ( -0.15%) 11878200.00 ( 12.55%) 13583200.00
length=16388: 13461700.00 ( -0.13%) 11558000.00 ( 14.03%) 13444100.00
length=16396: 13517500.00 ( -0.03%) 11561300.00 ( 14.45%) 13513900.00
length=16389: 13534100.00 ( 0.17%) 11756800.00 ( 13.28%) 13556900.00
length=16395: 13585600.00 ( 0.11%) 11791800.00 ( 13.30%) 13601200.00
length=16390: 13480100.00 ( -0.13%) 11685500.00 ( 13.20%) 13462100.00
length=16394: 13529900.00 ( -0.23%) 11549800.00 ( 14.43%) 13498200.00
length=16391: 13595400.00 ( -0.26%) 11768200.00 ( 13.22%) 13560600.00
length=16393: 13567000.00 ( 0.20%) 11779700.00 ( 13.35%) 13594700.00
length=32768: 71308800.00 ( -6.53%) 50220800.00 ( 24.98%) 66939200.00
length=32784: 72100800.00 (-11.55%) 50114100.00 ( 22.47%) 64636300.00
length=32769: 71767000.00 ( -7.10%) 51238400.00 ( 23.54%) 67010000.00
length=32783: 70113700.00 (-40.95%) 51129000.00 ( -2.78%) 49744400.00
length=32770: 71367600.00 ( -6.52%) 50244700.00 ( 25.01%) 67000900.00
length=32782: 64366700.00 ( 4.71%) 50101400.00 ( 25.83%) 67545600.00
length=32771: 71440100.00 ( -6.51%) 51263900.00 ( 23.57%) 67074900.00
length=32781: 66993000.00 ( 0.34%) 51108300.00 ( 23.97%) 67220300.00
length=32772: 71443900.00 (-60.50%) 50062100.00 (-12.47%) 44512600.00
length=32780: 71759100.00 ( -6.58%) 50263200.00 ( 25.35%) 67328600.00
length=32773: 71714900.00 (-33.21%) 51076600.00 ( 5.12%) 53835400.00
length=32779: 71756900.00 ( -6.56%) 51290800.00 ( 23.83%) 67337800.00
length=32774: 59689300.00 (-34.55%) 50068400.00 (-12.86%) 44363300.00
length=32778: 71847500.00 (-18.20%) 50084100.00 ( 17.61%) 60786500.00
length=32775: 71599300.00 ( -6.54%) 51278200.00 ( 23.70%) 67204800.00
length=32777: 71862900.00 (-60.85%) 51094000.00 (-14.36%) 44677900.00
length=65536: 282848000.00 ( -6.60%) 199187000.00 ( 24.93%) 265325000.00
length=65552: 243285000.00 (-41.61%) 198512000.00 (-15.54%) 171805000.00
length=65537: 255415000.00 (-23.47%) 202499000.00 ( 2.11%) 206858000.00
length=65551: 280122000.00 (-62.95%) 203349000.00 (-18.29%) 171911000.00
length=65538: 283676000.00 (-14.46%) 198368000.00 ( 19.96%) 247848000.00
length=65550: 275566000.00 (-51.76%) 198494000.00 ( -9.31%) 181581000.00
length=65539: 283699000.00 ( -6.58%) 203453000.00 ( 23.57%) 266195000.00
length=65549: 286572000.00 ( -6.65%) 202607000.00 ( 24.60%) 268712000.00
length=65540: 283710000.00 ( -6.59%) 199161000.00 ( 25.17%) 266160000.00
length=65548: 237573000.00 ( 11.48%) 198462000.00 ( 26.06%) 268395000.00
length=65541: 284150000.00 ( -6.58%) 203273000.00 ( 23.75%) 266600000.00
length=65547: 286250000.00 ( -6.70%) 202594000.00 ( 24.48%) 268263000.00
length=65542: 284167000.00 ( -6.60%) 199122000.00 ( 25.31%) 266584000.00
length=65546: 285656000.00 ( -6.59%) 198443000.00 ( 25.95%) 268002000.00
length=65543: 284600000.00 ( -6.58%) 203247000.00 ( 23.89%) 267030000.00
length=65545: 285665000.00 ( -6.40%) 202575000.00 ( 24.55%) 268472000.00
<snip>
* sysdeps/aarch64/multiarch/Makefile (sysdep_routines): Add
memmove_falkor.
* sysdeps/aarch64/multiarch/ifunc-impl-list.c
(__libc_ifunc_impl_list): Likewise.
* sysdeps/aarch64/multiarch/memmove.c: Likewise.
* sysdeps/aarch64/multiarch/memmove_falkor.S: New file.
glibc has an add-ons mechanism to allow additional software to be
integrated into the glibc build. Such add-ons may be within the glibc
source tree, or outside it at a path passed to the --enable-add-ons
configure option.
localedata and crypt were once add-ons, distributed in separate
release tarballs, but long since stopped using that mechanism.
Linuxthreads was always an add-on. Ports spent some time as an add-on
with separate release tarballs, then was first moved into the glibc
source tree, then had its sysdeps files moved into the main sysdeps
hierarchy so the add-ons mechanism was no longer used. NPTL spent
some time as an add-on in the main glibc tree before stopping using
the add-on mechanism. libidn used to have separate release tarballs
but no longer does so, but still uses the add-ons mechanism within the
glibc source tree. Various other software has supported building with
the add-ons mechanism at times in the past, but I don't think any is
still widely used.
Add-ons involve significant, little-used complexity in the glibc build
system, and make it hard to understand what the space of possible
glibc configurations is. This patch removes the add-ons mechanism.
libidn is now built via the Subdirs mechanism to cause any
configuration using sysdeps/unix/inet to build libidn; HAVE_LIBIDN
(which effectively means shared libraries are available) is now
defined via sysdeps/unix/inet/configure. Various references to
add-ons around the source tree are removed (in the case of maint.texi,
the example list of sysdeps directories is still very out of date).
Externally maintained ports should now put their files in the normal
sysdeps directory structure rather than being arranged as add-ons;
they probably need to change e.g. elf.h anyway, rather than actually
being able to work just as a drop-in subtree. Hurd libpthread should
be arranged similarly to NPTL, so some files might go in a
hurd-pthreads (or similar) top-level directory in glibc, while sysdeps
files should go in the normal sysdeps directory structure (possibly in
hurd or hurd-pthreads subdirectories, just as there are nptl
subdirectories in the sysdeps tree).
Tested for x86_64, and with build-many-glibcs.py.
* configure.ac (--enable-add-ons): Remove option.
(machine): Do not mention add-ons in comment.
(LIBC_PRECONFIGURE): Likewise.
(add_ons): Remove variable and sanity checks and logic to locate
add-ons.
(add_ons_automatic): Remove variable.
(configured_add_ons): Likewise.
(add_ons_sfx): Likewise.
(add_ons_pfx): Likewise.
(add_on_subdirs): Likewise.
(sysnames_add_ons): Likewise. Remove loop over add-ons and
consideration of add-ons in Implies handling.
(sysdeps_add_ons): Likewise.
* configure: Regenerated.
* libidn/configure.ac: Remove.
* libidn/configure: Likewise.
* sysdeps/unix/inet/configure.ac: New file.
* sysdeps/unix/inet/configure: New generated file.
* sysdeps/unix/inet/Subdirs: Add libidn.
* Makeconfig (sysdeps-srcdirs): Remove variable.
(+sysdep_dirs): Do not include $(sysdeps-srcdirs).
($(common-objpfx)config.status): Do not depend on add-on files.
($(common-objpfx)shlib-versions.v.i): Do not mention add-ons in
comment.
(all-subdirs): Do not include $(add-on-subdirs).
* Makefile (dist-prepare): Do not use $(sysdeps-add-ons).
* config.make.in (add-ons): Remove variable.
(add-on-subdirs): Likewise.
(sysdeps-add-ons): Likewise.
* manual/Makefile (add-chapters): Remove.
($(objpfx)texis): Do not depend on $(add-chapters).
(nonexamples): Do not handle $(add-chapters).
(examples): Do not handle $(add-ons).
(chapters.% top-menu.%): Do not pass '$(add-chapters)' to
libc-texinfo.sh.
* manual/install.texi (Installation): Do not mention add-ons.
(--enable-add-ons): Do not document configure option.
* INSTALL: Regenerated.
* manual/libc-texinfo.sh: Do not handle $2 add-ons argument.
* manual/maint.texi (Hierarchy Conventions): Do not mention
add-ons.
* scripts/build-many-glibcs.py (Glibc.build_glibc): Do not use
--enable-add-ons.
* scripts/gen-sorted.awk: Do not handle Subdirs files from
add-ons.
* scripts/test-installation.pl: Do not handle glibc-compat add-on.
* sysdeps/nptl/Makeconfig: Do not mention add-ons in comment.
On i386, when multi-arch is enabled, all external functions must be
called via PIC PLT in PIE, which requires setting up EBX register,
since they may be IFUNC functions.
* config.h.in (NO_HIDDEN_EXTERN_FUNC_IN_PIE): New.
* include/libc-symbols.h (__hidden_proto_hiddenattr): Add check
for PIC and NO_HIDDEN_EXTERN_FUNC_IN_PIE.
* sysdeps/i386/configure.ac (NO_HIDDEN_EXTERN_FUNC_IN_PIE): New
AC_DEFINE if multi-arch is enabled.
* sysdeps/i386/configure: Regenerated.
This patch makes dbl-64 fma use libm_alias_double. The ldbl-opt
version is removed. The sparc32 version no longer needs to handle
compat symbols, while alpha needs a new wrapper to avoid getting the
ldbl-128 version (where ldbl-opt is earlier in the list of sysdeps
directories, so previously fma came from there).
Tested for x86_64, and tested with build-many-glibcs.py that installed
stripped shared libraries are unchanged by the patch.
* sysdeps/ieee754/dbl-64/s_fma.c: Include <libm-alias-double.h>.
(fma): Define using libm_alias_double.
* sysdeps/ieee754/ldbl-opt/s_fma.c: Remove file.
* sysdeps/sparc/sparc32/fpu/s_fma.c: Do not include
<math_ldbl_opt.h>.
(fmal): Do not define as compat symbol here.
* sysdeps/alpha/fpu/s_fma.c: New file.
32-bit SPARC libm should have compat symbols for copysignl
(GLIBC_2.0), fabsl (GLIBC_2.0), fmal (GLIBC_2.1), pointing to the
double functions; they were present in glibc 2.8, for example, but are
now missing, probably when optimized SPARC function implementations
were added without appropriate compat symbol handling. The same
applies to copysignl in libc. This patch restores those compat
symbols.
Tested with build-many-glibcs.py for sparcv9-linux-gnu.
[BZ #22229]
* sysdeps/sparc/sparc32/fpu/s_copysign.S: Include
<math_ldbl_opt.h>
(copysignl): Define as compat symbol at version GLIBC_2_0 for libm
and libc.
* sysdeps/sparc/sparc32/fpu/s_fabs.S: Include <math_ldbl_opt.h>.
(fabsl): Define as compat symbol at version GLIBC_2_0 for libm.
* sysdeps/sparc/sparc32/fpu/s_fma.c: Include <math_ldbl_opt.h>.
(fmal): Define as compat symbol at version GLIBC_2_1 for libm.
* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_copysign.S:
Include <math_ldbl_opt.h>
(copysignl): Define as compat symbol at version GLIBC_2_0 for libm
and libc.
(compat_symbol): Undefine and redefine.
* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fabs.S: Include
<math_ldbl_opt.h>
(fabsl): Define as compat symbol at version GLIBC_2_0 for libm.
(compat_symbol): Undefine and redefine.
* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fma.c
[HAVE_AS_VIS3_SUPPORT]: Include <math_ldbl_opt.h>.
[HAVE_AS_VIS3_SUPPORT] (fmal): Define as compat symbol at version
GLIBC_2_1 for libm.
* sysdeps/unix/sysv/linux/sparc/sparc32/libc.abilist: Add
GLIBC_2.0 copysignl symbol.
* sysdeps/unix/sysv/linux/sparc/sparc32/libm.abilist: Add
GLIBC_2.0 copysignl and fabsl and GLIBC_2.1 fmal symbols.
Given my recent changes, sysdeps/alpha/fpu/s_nearbyint.c is no longer
needed: it just includes the dbl-64/wordsize-64 version, which is the
one that would be used anyway, and defines a compat symbol,
duplicating the same compat symbol defined by the dbl-64/wordsize-64
version through use of libm_alias_double. Thus, this patch removes
the redundant wrapper.
Tested with build-many-glibcs.py that installed stripped shared
libraries are unchanged for alpha.
* sysdeps/alpha/fpu/s_nearbyint.c: Remove file.
Without SVID compat wrapper yn(n,0) and ynf(n,0) does not raise
the divide-by-zero excpetion and it may return inf with the wrong
sign for n < 0.
[BZ #22244]
* sysdeps/ieee754/dbl-64/e_jn.c (__ieee754_yn): Fix x == 0 case.
* sysdeps/ieee754/flt-32/e_jnf.c (__ieee754_ynf): Likewise.
On 64bit targets if the SVID compat wrapper is suppressed (e.g. static linking)
then log2(0) and log10(0) returned inf instead of -inf.
[BZ #22243]
* sysdeps/ieee754/dbl-64/wordsize-64/e_log10.c (__ieee754_log10): Use fabs.
* sysdeps/ieee754/dbl-64/wordsize-64/e_log2.c (__ieee754_log2): Likewise.
Don't use "leal main@GOTOFF(%ebx), %eax" since main may be in a
shared object. Linker will convert "movl main@GOT(%ebx), %eax"
to "leal main@GOTOFF(%ebx), %eax" if main is defined locally.
* sysdeps/i386/start.S: Replace "leal main@GOT(%ebx), %eax" with
"movl main@GOTOFF(%ebx), %eax".
This code is used in non-PIE static executable and static PIE. It checks
if _DYNAMIC is undefined before using it to compute load address. But
not all targets can convert access _DYNAMIC via GOT, which needs dynamic
relocation, to PC-relative at link-time.
* sysdeps/i386/dl-machine.h (elf_machine_load_address): Don't
allow undefined _DYNAMIC in PIE libc.a.
* sysdeps/x86_64/dl-machine.h (elf_machine_load_address):
Likewse.
Since mips can't convert access _DYNAMIC via GOT, which needs dynamic
relocation, to PC-relative at link-time, don't check _DYNAMIC in
elf_machine_load_address.
* sysdeps/mips/dl-machine.h (elf_machine_load_address): Don't
check _DYNAMIC.
Since arm can't convert access _DYNAMIC via GOT, which needs dynamic
relocation, to PC-relative at link-time, don't check _DYNAMIC in
elf_machine_load_address.
* sysdeps/arm/dl-machine.h (elf_machine_load_address): Don't
check _DYNAMIC.
This patch makes dbl-64 modf use libm_alias_double. Both the dbl-64
and dbl-64/wordsize-64 versions are changed, and the ldbl-opt version
is changed to define the libc compat symbol only. Because of
multiarch wrappers, the changed implementations are made not to define
aliases at all if __modf is defined as a macro, as with other
functions, so avoiding duplicate compat symbols while allowing those
wrappers to be simplified.
Tested for x86_64, and verified with build-many-glibcs.py that
installed stripped shared libraries are unchanged by the patch.
* sysdeps/ieee754/dbl-64/s_modf.c: Include <libm-alias-double.h>.
(modf): Define using libm_alias_double, only if [!__modf].
* sysdeps/ieee754/dbl-64/wordsize-64/s_modf.c: Include
<libm-alias-double.h>.
(modf): Define using libm_alias_double, only if [!__modf].
* sysdeps/ieee754/ldbl-opt/s_modf.c (modfl): Only define libc
compat symbol here.
* sysdeps/powerpc/powerpc32/power4/fpu/multiarch/s_modf-ppc32.c
(weak_alias): Do not undefine and redefine.
(strong_alias): Likewise.
* sysdeps/powerpc/powerpc64/fpu/multiarch/s_modf-ppc64.c
(weak_alias): Likewise.
(strong_alias): Likewise.
This patch makes dbl-64 logb use libm_alias_double. Both the dbl-64
and dbl-64/wordsize-64 versions are changed, and the ldbl-opt version
is removed. Because of multiarch wrappers, the changed
implementations are made not to define aliases at all if __logb is
defined as a macro, as with other functions, so avoiding duplicate
compat symbols while allowing those wrappers to be simplified.
Tested for x86_64, and verified with build-many-glibcs.py that
installed stripped shared libraries are unchanged (except on alpha
where changes from using the wordsize-64 version are expected).
* sysdeps/ieee754/dbl-64/s_logb.c: Include <libm-alias-double.h>.
(logb): Define using libm_alias_double, only if [!__logb].
* sysdeps/ieee754/dbl-64/wordsize-64/s_logb.c: Include
<libm-alias-double.h>.
(logb): Define using libm_alias_double, only if [!__logb].
* sysdeps/ieee754/ldbl-opt/s_logb.c: Remove file.
* sysdeps/powerpc/powerpc32/power4/fpu/multiarch/s_logb-ppc32.c
(weak_alias): Do not undefine and redefine.
(strong_alias): Likewise.
* sysdeps/powerpc/powerpc64/fpu/multiarch/s_logb-ppc64.c
(weak_alias): Likewise.
(strong_alias): Likewise.
For static PIE code, PIC is defined and SHARED is undefined. We
should check SHARED instead PIC for SYSCALL_ERROR_NAME.
* sysdeps/unix/sysv/linux/tile/sysdep.h (SYSCALL_ERROR_NAME):
Check SHARED instead PIC.
This patch makes the implementation of fmaf in the dbl-64 directory
use libm_alias float.
Tested for x86_64, and verified with build-many-glibcs.py that
installed stripped shared libraries are unchanged by this patch.
* sysdeps/ieee754/dbl-64/s_fmaf.c: Include <libm-alias-float.h>.
[!__fmaf] (fmaf): Define using libm_alias_float.
This patch makes dbl-64 frexp use libm_alias_double. Both the dbl-64
and dbl-64/wordsize-64 versions are changed; the ldbl-opt version is
made to define only the libc frexpl compat symbol, now the generic
code handles the libm compat symbol automatically.
Tested for x86_64, and verified with build-many-glibcs.py that
installed stripped shared libraries are unchanged by this patch.
* sysdeps/ieee754/dbl-64/s_frexp.c: Include <libm-alias-double.h>.
(frexp): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/wordsize-64/s_frexp.c: Include
<libm-alias-double.h>.
(frexp): Define using libm_alias_double.
* sysdeps/ieee754/ldbl-opt/s_frexp.c (frexpl): Only define libc
compat symbol here.
All representations of floating-point numbers in types with IEC 60559
binary exchange format are canonical. On the other hand, types with IEC
60559 extended formats, such as those implemented under ldbl-96 and
ldbl-128ibm, contain representations that are not canonical.
TS 18661-1 introduced the type-generic macro iscanonical, which returns
whether a floating-point value is canonical or not. In Glibc, this
type-generic macro is implemented using the macro __MATH_TG, which, when
support for float128 is enabled, relies on __builtin_types_compatible_p
to select between floating-point types. However, this use of
iscanonical breaks C++ applications, because the builtin is only
available in C mode.
This patch provides a C++ implementation of iscanonical that relies on
function overloading, rather than builtins, to select between
floating-point types.
Unlike the C++ implementations for iszero and issignaling, this
implementation ignores __NO_LONG_DOUBLE_MATH. The double type always
matches IEC 60559 double format, which is always canonical. Thus, when
double and long double are the same (__NO_LONG_DOUBLE_MATH), iscanonical
always returns 1 and is not implemented with __MATH_TG.
Tested for powerpc64, powerpc64le and x86_64.
[BZ #22235]
* math/math.h: Trivial fix for unbalanced parentheses in comment.
* math/Makefile [CXX] (tests): Add test-math-iscanonical.cc.
(CFLAGS-test-math-iscanonical.cc): New variable.
* math/test-math-iscanonical.cc: New file.
* sysdeps/ieee754/ldbl-96/bits/iscanonical.h (iscanonical):
Provide a C++ implementation based on function overloading,
rather than using __MATH_TG, which uses C-only builtins.
* sysdeps/ieee754/ldbl-128ibm/bits/iscanonical.h (iscanonical):
Likewise.
* sysdeps/powerpc/powerpc64le/Makefile
(CFLAGS-test-math-iscanonical.cc): New variable.
This patch makes more dbl-64 functions use libm_alias_double to define
function aliases. Specifically, it makes the change for functions
with dbl-64/wordsize-64 versions, changing both the dbl-64 and
dbl-64/wordsize-64 versions and removing the ldbl-opt wrappers.
Functions are excluded from this patch if there are complications
because of versions of those functions also present in libc, or
architecture-specific wrappers round these files.
Tested for x86_64, and with build-many-glibcs.py. Installed stripped
shared libraries are unchanged except for alpha (where increased use
of dbl-64/wordsize-64 files, where previously ldbl-opt files that
wrapped dbl-64 files were used, was expected to result in different,
better code).
* sysdeps/ieee754/dbl-64/s_ceil.c: Include <libm-alias-double.h>.
(ceil): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/s_floor.c: Include <libm-alias-double.h>.
(floor): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/s_llround.c: Include
<libm-alias-double.h>.
(llround): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/s_lround.c: Include
<libm-alias-double.h>.
(lround): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/s_nearbyint.c: Include
<libm-alias-double.h>.
(nearbyint): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/s_remquo.c: Include
<libm-alias-double.h>.
(remquo): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/s_rint.c: Include <libm-alias-double.h>.
(rint): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/s_round.c: Include <libm-alias-double.h>.
(round): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/s_trunc.c: Include <libm-alias-double.h>.
(trunc): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/wordsize-64/s_ceil.c: Include
<libm-alias-double.h>.
(ceil): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/wordsize-64/s_floor.c: Include
<libm-alias-double.h>.
(floor): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/wordsize-64/s_llround.c: Include
<libm-alias-double.h>.
(llround): Define using libm_alias_double.
[_LP64] (lround): Likewise.
* sysdeps/ieee754/dbl-64/wordsize-64/s_lround.c: Include
<libm-alias-double.h>.
[!_LP64] (lround): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/wordsize-64/s_nearbyint.c: Include
<libm-alias-double.h>.
(nearbyint): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/wordsize-64/s_remquo.c: Include
<libm-alias-double.h>.
(remquo): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/wordsize-64/s_rint.c: Include
<libm-alias-double.h>.
(rint): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/wordsize-64/s_round.c: Include
<libm-alias-double.h>.
(round): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/wordsize-64/s_trunc.c: Include
<libm-alias-double.h>.
(trunc): Define using libm_alias_double.
* sysdeps/ieee754/ldbl-opt/s_ceil.c: Remove file.
* sysdeps/ieee754/ldbl-opt/s_floor.c: Likewise.
* sysdeps/ieee754/ldbl-opt/s_llround.c: Likewise.
* sysdeps/ieee754/ldbl-opt/s_lround.c: Likewise.
* sysdeps/ieee754/ldbl-opt/s_nearbyint.c: Likewise.
* sysdeps/ieee754/ldbl-opt/s_remquo.c: Likewise.
* sysdeps/ieee754/ldbl-opt/s_rint.c: Likewise.
* sysdeps/ieee754/ldbl-opt/s_round.c: Likewise.
* sysdeps/ieee754/ldbl-opt/s_trunc.c: Likewise.
(&_dl_main_map) is used instead of (&bootstrap_map) to bootstrap static
PIE. Define BOOTSTRAP_MAP with (&_dl_main_map) to avoid hardcode to
(&bootstrap_map).
* elf/rtld.c (BOOTSTRAP_MAP): New.
(RESOLVE_MAP): Replace (&bootstrap_map) with BOOTSTRAP_MAP.
* sysdeps/hppa/dl-machine.h (ELF_MACHINE_BEFORE_RTLD_RELOC):
Likewise.
* sysdeps/ia64/dl-machine.h (ELF_MACHINE_BEFORE_RTLD_RELOC):
Likewise.
* sysdeps/mips/dl-machine.h (ELF_MACHINE_BEFORE_RTLD_RELOC):
Likewise.
On the Hurd, the rtld needs to see its own dumb versions of a few functions
(defined in sysdeps/mach/hurd/dl-sysdep.c) overridden by libc's versions once
loaded. rtld should thus not have hidden attribute for these. To achieve this,
the Hurd port used to just define NO_HIDDEN, which disables it completely. For
now, this changes that to disabling it for all rtld functions, for simplicity.
See Roland's comment on https://sourceware.org/bugzilla/show_bug.cgi?id=15605#c5
The ld.so numbers remain at
8 .rel.plt 000000c8 00000c24 00000c24 00000c24 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
9 .plt 000001a0 00000cf0 00000cf0 00000cf0 2**4
CONTENTS, ALLOC, LOAD, READONLY, CODE
10 .plt.got 00000010 00000e90 00000e90 00000e90 2**3
CONTENTS, ALLOC, LOAD, READONLY, CODE
18 .got.plt 00000070 0002d000 0002d000 0002c000 2**2
CONTENTS, ALLOC, LOAD, DATA
which is about 3 times as much as on Linux.
The libc.so numbers get divided by 3 (the remainings are mostly RPC stub calls)
* include/libc-symbols.h [NO_RTLD_HIDDEN] (rtld_hidden_proto,
rtld_hidden_tls_proto, rtld_hidden_def, rtld_hidden_weak,
rtld_hidden_rtld_hidden_ver, data_def, rtld_hidden_data_weak,
rtld_hidden_data_ver): Define to empty.
* include/assert.h [IS_IN(rtld) && NO_RTLD_HIDDEN] (__assert_fail,
__assert_perror_fail): Likewise.
* include/dirent.h [IS_IN(rtld) && NO_RTLD_HIDDEN]
(__rewinddir): Likewise.
* include/libc-internal.h [IS_IN(rtld) && NO_RTLD_HIDDEN]
(__profile_frequency): Likewise.
* include/setjmp.h (__sigsetjmp): Likewise.
* include/signal.h [IS_IN(rtld) && NO_RTLD_HIDDEN] (__sigaction,
__libc_sigaction): Likewise.
* include/stdlib.h [NO_RTLD_HIDDEN] (unsetenv, __strtoul_internal): Do
not set hidden attribute.
* include/string.h [IS_IN(rtld) && NO_RTLD_HIDDEN] (__stpcpy, __strdup,
__strerror_t, __strsep_g, memchr, memcmp, memcpy, memmove, memset,
rawmemchr, stpcpy, strchr, strcmp, strlen, strnlen, strsep): Likewise.
* include/sys/stat.h [IS_IN(rtld) && NO_RTLD_HIDDEN] (__fxstat,
__fxstat64, __lxstat, __lxstat64, __xstat, __xstat64,
__fxstatat64): Likewise.
* include/sys/utsname.h [IS_IN(rtld) && NO_RTLD_HIDDEN]
(__uname): Likewise.
* include/sysdeps/generic/_itoa.h [IS_IN(rtld) && NO_RTLD_HIDDEN]
(_itoa_upper_digits, _itoa_lower_digits): Likewise.
* sysdeps/mach/hurd/configure.ac (NO_HIDDEN): Do not set.
(NO_RTLD_HIDDEN): Set.
* sysdeps/mach/hurd/configure: Refresh.
* config.h.in: Refresh.
This patch makes the dbl-64 atan and tan implementations use
libm_alias_double, removing the corresponding ldbl-opt wrappers.
Tested for x86_64, and with build-many-glibcs.py. Installed stripped
shared libraries are unchanged on non-ldbl-opt platforms. For
ldbl-opt configurations, the patch has the effect of causing
compat_symbol to define atanl and tanl in terms of __atan and __tan
instead of in terms of atan and tan, which is enough to change the
installed stripped libm.so.
* sysdeps/ieee754/dbl-64/s_atan.c: Include <libm-alias-double.h>.
(atan): Define using libm_alias_double.
* sysdeps/ieee754/dbl-64/s_tan.c: Include <libm-alias-double.h>.
(tan): Define using libm_alias_double.
* sysdeps/ieee754/ldbl-opt/s_atan.c: Remove file.
* sysdeps/ieee754/ldbl-opt/s_tan.c: Likewise.
This patch converts the dbl-64 implementations of atan and tan into
weak aliases of __atan and __tan, in preparation for making them use
libm_alias_double. Consequent changes are made to the x86_64
multiarch versions wrapping round them (with the dbl-64 functions,
like other such functions, being made not to define their aliases at
all if __atan or __tan are defined as macros by an including file).
Tested for x86_64, and with build-many-glibcs.py.
* sysdeps/ieee754/dbl-64/s_atan.c (atan): Rename to __atan and
define as weak alias of __atan. Do not define any aliases if
[__atan].
[NO_LONG_DOUBLE] (__atanl): Define as strong alias of __atan.
[NO_LONG_DOUBLE] (atanl): Define as weak alias of __atanl.
* sysdeps/ieee754/dbl-64/s_tan.c (tan): Rename to __tan and define
as weak alias of __tan. Do not define any aliases if [__tan].
[NO_LONG_DOUBLE] (__tanl): Define as strong alias of __tan.
[NO_LONG_DOUBLE] (tanl): Define as weak alias of __tanl.
* sysdeps/x86_64/fpu/multiarch/s_atan-avx.c (atan): Rename to
__atan.
* sysdeps/x86_64/fpu/multiarch/s_atan-fma.c (atan): Likewise.
* sysdeps/x86_64/fpu/multiarch/s_atan-fma4.c (atan): Likewise.
* sysdeps/x86_64/fpu/multiarch/s_atan.c (atan): Rename to __atan
and define as weak alias of __atan.
* sysdeps/x86_64/fpu/multiarch/s_tan-avx.c (tan): Rename to
__atan.
* sysdeps/x86_64/fpu/multiarch/s_tan-fma.c (tan): Likewise.
* sysdeps/x86_64/fpu/multiarch/s_tan-fma4.c (tan): Likewise.
* sysdeps/x86_64/fpu/multiarch/s_tan.c (tan): Rename to __tan and
define as weak alias of __tan.
The new generic logf, log2f and powf code don't need wrappers any more,
they set errno inline so only use the wrappers on targets that need it.
* sysdeps/ieee754/flt-32/e_log2f.c (__log2f): Define without wrapper.
* sysdeps/ieee754/flt-32/e_logf.c (__logf): Likewise
* sysdeps/ieee754/flt-32/e_powf.c (__powf): Likewise
* sysdeps/ieee754/flt-32/w_log2f.c: New file.
* sysdeps/ieee754/flt-32/w_logf.c: New file.
* sysdeps/ieee754/flt-32/w_powf.c: New file.
* sysdeps/i386/fpu/w_log2f.c: New file.
* sysdeps/i386/fpu/w_logf.c: New file.
* sysdeps/i386/fpu/w_powf.c: New file.
* sysdeps/m68k/m680x0/fpu/w_log2f.c: New file.
* sysdeps/m68k/m680x0/fpu/w_logf.c: New file.
* sysdeps/m68k/m680x0/fpu/w_powf.c: New file.
The new generic expf and exp2f code don't need wrappers any more, they
set errno inline, so only use the wrappers on targets that need it.
(If the wrapper is needed, then the top level wrapper code is included,
otherwise empty w_exp*f.c is used to suppress the wrapper.)
A powerpc64 expf implementation includes the expf c code directly which
needed some changes.
* sysdeps/ieee754/flt-32/e_exp2f.c (__exp2f): Define without wrapper.
* sysdeps/ieee754/flt-32/e_expf.c (__expf): Likewise
* sysdeps/ieee754/flt-32/w_exp2f.c: New file.
* sysdeps/ieee754/flt-32/w_expf.c: New file.
* sysdeps/powerpc/powerpc64/fpu/multiarch/e_expf-ppc64.c: Update for
the new expf code.
* sysdeps/powerpc/powerpc64/fpu/multiarch/w_expf.c: New file.
* sysdeps/powerpc/powerpc64/power8/fpu/w_expf.c: New file.
* sysdeps/m68k/m680x0/fpu/w_exp2f.c: New file.
* sysdeps/m68k/m680x0/fpu/w_expf.c: New file.
* sysdeps/i386/fpu/w_exp2f.c: New file.
* sysdeps/i386/fpu/w_expf.c: New file.
* sysdeps/i386/i686/fpu/multiarch/w_expf.c: New file.
* sysdeps/x86_64/fpu/w_expf.c: New file.
