I like the idea.. a processor inside a memory

Will In-Memory Processing Work?

Changes that sidestep von Neumann architecture could be key to low-power ML hardware.

@nm so basically a cluster of CPUs with enormous L1, L2 and L3 cache?

@nm a very good question, as I'm not a CPU architect.

@matt @nm or maybe a cluster of memory with l1, l2 and l3 computing units

This idea has been around for at least 10 years if not 20. It couldn't justify the added complexity back then and as far as I am aware of, it still cannot. Keep in mind, this is not a minor change in the computer architecture, but a major change in the paradigm. Hence requires different programming models. And the speed gains are not as large as promised for most workloads.

Sign in to participate in the conversation

Linux geeks doing what Linux geeks do...