Good slides. I didn't know about the potential for auto-parallelizing vector operations. This should allow me to eliminate my boilerplate chunk/send to worker/reassemble code.