I agree. I've had to wrap non-amd libs far too many times.

I wonder how this will work when optimized.