Hi,
I've finished ARM NEON SIMD optimization for 44 kernels which means
only 23 kernels (not often used) are now missing a NEON
implementation.
All of the kernels so far are written using compiler intrinsics which
means they will compile on both arm 32 and 64 bit. I've experimented
with aarch64 assembly ("neonv8") but found that in practice it's very
difficult to get any meaningful performance gain. Gcc is pretty damn
good at generating code and scheduling instructions...
Issue with link to my branch is here:
https://github.com/gnuradio/volk/issues/243
--Albin
_______________________________________________
Discuss-gnuradio mailing list
Discuss-gnuradio@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
