vitalif
/
gf-complete
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

configure: add ARM/AArch64 NEON support 

Checks for arm_neon.h header.
master
Janne Grunau 2014-09-04 18:29:58 +02:00
parent 568df90edc
commit eb5ce0ca42
3 changed files with 74 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
configure.ac
View File

@ -24,6 +24,27 @@ AC_PROG_CC
AX_EXT()
AC_ARG_ENABLE([neon],
AS_HELP_STRING([--disable-neon], [Build without NEON optimizations]))
AS_IF([test "x$enable_neon" != "xno"],
[noneon_CPPFLAGS=$CPPFLAGS
CPPFLAGS="$CPPFLAGS $SIMD_FLAGS"
AC_CHECK_HEADER([arm_neon.h],
[have_neon=yes],
[have_neon=no
CPPFLAGS=$noneon_CPPFLAGS])],
[have_neon=no
AS_IF([test "x$ax_cv_have_neon_ext" = "xyes"],
[SIMD_FLAGS=""])
])
AS_IF([test "x$have_neon" = "xno"],
[AS_IF([test "x$enable_neon" = "xyes"],
[AC_MSG_ERROR([neon requested but arm_neon.h not found])])
])
AM_CONDITIONAL([HAVE_NEON], [test "x$have_neon" = "xyes"])
AC_ARG_ENABLE([sse],
AS_HELP_STRING([--disable-sse], [Build without SSE optimizations]),
[if test "x$enableval" = "xno" ; then

4
include/gf_complete.h
View File

@ -33,6 +33,10 @@
#include <wmmintrin.h>
#endif
#if defined(ARM_NEON)
#include <arm_neon.h>
#endif
/* These are the different ways to perform multiplication.
Not all are implemented for all values of w.

49
m4/ax_ext.m4
View File

@ -41,6 +41,55 @@ AC_DEFUN([AX_EXT],
AC_REQUIRE([AC_CANONICAL_HOST])
case $host_cpu in
aarch64*)
AC_DEFINE(HAVE_ARCH_AARCH64,,[targeting AArch64])
SIMD_FLAGS="$SIMD_FLAGS -DARCH_AARCH64"
AC_CACHE_CHECK([whether NEON is supported], [ax_cv_have_neon_ext],
[
# TODO: detect / cross-compile
ax_cv_have_neon_ext=yes
])
AC_CACHE_CHECK([whether cryptographic extension is supported], [ax_cv_have_arm_crypt_ext],
[
# TODO: detect / cross-compile
ax_cv_have_arm_crypt_ext=yes
])
if test "$ax_cv_have_arm_crypt_ext" = yes; then
AC_DEFINE(HAVE_ARM_CRYPT_EXT,,[Support ARM cryptographic extension])
fi
if test "$ax_cv_have_neon_ext" = yes; then
AC_DEFINE(HAVE_NEON,,[Support NEON instructions])
fi
if test "$ax_cv_have_arm_crypt_ext" = yes && test "$ax_cv_have_neon_ext" = yes; then
AX_CHECK_COMPILE_FLAG(-march=armv8-a+simd+crypto,
SIMD_FLAGS="$SIMD_FLAGS -march=armv8-a+simd+crypto -DARM_CRYPT -DARM_NEON", [])
elif test "$ax_cv_have_arm_crypt_ext" = yes; then
AX_CHECK_COMPILE_FLAG(-march=armv8-a+crypto,
SIMD_FLAGS="$SIMD_FLAGS -march=armv8-a+crypto -DARM_CRYPT", [])
elif test "$ax_cv_have_neon_ext" = yes; then
AX_CHECK_COMPILE_FLAG(-march=armv8-a+simd,
SIMD_FLAGS="$SIMD_FLAGS -march=armv8-a+simd -DARM_NEON", [])
fi
;;
arm*)
AC_CACHE_CHECK([whether NEON is supported], [ax_cv_have_neon_ext],
[
# TODO: detect / cross-compile
ax_cv_have_neon_ext=yes
])
if test "$ax_cv_have_neon_ext" = yes; then
AC_DEFINE(HAVE_NEON,,[Support NEON instructions])
AX_CHECK_COMPILE_FLAG(-mfpu=neon,
SIMD_FLAGS="$SIMD_FLAGS -mfpu=neon -DARM_NEON", [])
fi
;;
powerpc*)
AC_CACHE_CHECK([whether altivec is supported], [ax_cv_have_altivec_ext],
[