Highly parallel workloads are an important niche in computing.
On Mon, 10 June 2024, 8:48 am Scott Baker via cctalk, <cctalk(a)classiccmp.org>
wrote:
I think the biggest change is our compute resources
stopped going faster
in terms of raw cycles per second, and started going wider in terms of
parallelism. It's now commonplace for me to run workloads that can actually
use many CPU cores, and I'm starting to occasionally run workloads that are
so parallel, that a GPU is a more suitable resource. At the same time as
the surge in parallelism, there's also a focus on going greener. I think
the last couple years have been particularly transformative.
Scott