Transceiver52M: Add ARM NEON support
Similar to the existing Intel SSE cases, add support for NEON vector
floating point SIMD processing. In this case, use ARM assembly
directly as the NEON intrinsics do not generate preferential code
output.
Currently support NEON vectorized convolution and floating point
integer conversions.
Signed-off-by: Thomas Tsou <tom@tsou.cc>
diff --git a/configure.ac b/configure.ac
index 848f887..ae4ea3b 100644
--- a/configure.ac
+++ b/configure.ac
@@ -68,6 +68,25 @@
[enable single daughterboard use on USRP1])
])
+AC_ARG_WITH(neon, [
+ AS_HELP_STRING([--with-neon],
+ [enable ARM NEON support])
+])
+
+AC_ARG_WITH(neon-vfpv4, [
+ AS_HELP_STRING([--with-neon-vfpv4],
+ [enable ARM NEON FMA support])
+])
+
+AS_IF([test "x$with_neon" = "xyes"], [
+ AC_DEFINE(HAVE_NEON, 1, Support ARM NEON)
+])
+
+AS_IF([test "x$with_neon_vfpv4" = "xyes"], [
+ AC_DEFINE(HAVE_NEON, 1, Support ARM NEON)
+ AC_DEFINE(HAVE_NEON_FMA, 1, Support ARM NEON with FMA)
+])
+
AS_IF([test "x$with_usrp1" = "xyes"], [
PKG_CHECK_MODULES(USRP, usrp >= 3.3)
])
@@ -85,6 +104,8 @@
AX_EXT
AM_CONDITIONAL(USRP1, [test "x$with_usrp1" = "xyes"])
+AM_CONDITIONAL(ARCH_ARM, [test "x$with_neon" = "xyes" || test "x$with_neon_vfpv4" = "xyes"])
+AM_CONDITIONAL(ARCH_ARM_A15, [test "x$with_neon_vfpv4" = "xyes"])
PKG_CHECK_MODULES(LIBUSB, libusb-1.0)
@@ -94,8 +115,9 @@
CommonLibs/Makefile \
GSM/Makefile \
Transceiver52M/Makefile \
+ Transceiver52M/arm/Makefile \
Transceiver52M/x86/Makefile \
sqlite3/Makefile \
-])
+])
AC_OUTPUT