Vectorization #596
Description
It seems that some lvl1 kernels (xdot, xasum, xnrm2, etc.) don't use vectorization. It would probably greatly increase the bandwidth achieved by those functions, is it something to add?
{{ message }}
It seems that some lvl1 kernels (xdot, xasum, xnrm2, etc.) don't use vectorization. It would probably greatly increase the bandwidth achieved by those functions, is it something to add?