Does anyone know if there are binaries available of ARM's OpenMax DL
implementation? It's written in hand-optimized neon assembly and the
source is available from this link: http://www.arm.com/community/multimedia/standards-apis.php
(Direct link is here: https://silver.arm.com/browse/OX002 but requires
free registration). I downloaded the code, but it appears to require
ARM's proprietary RVDS compiler.
Also, does anyone have any insights into the performance of the GCC
neon intrinsics compared to OpenMax DL? I realize this is probably
mostly algorithm dependent, but it'd be nice if someone has some high
level ideas. Searching through some of the other neon related posts,
it seems clear that performance is highly implementation dependent.