Can someone clarify what kinds of systems are better at blocked assignment vs interleaved assignment?

On slide 33 it mentions that interleaved assignment requires much more inter-process communication than blocked assignment. On systems that aren't optimized for inter-thread communication, the interleaved version would be much slower. However, if a system was optimized for inter-thread communication there would be a case for interleaved assignment. For example, if a system had support for many threads on a single core and sharing a single cache then it could allow for relatively fast inter-thread communication.

I would also suggest that blocked assignment might be more cache friendly in some cases.