News for hackers. Stuff that matters.
Re:U.S. Focuses on Invigorating Chiplets' to Stay Cuting- Ed... by bee_rider on Friday May 12, 2023 @06:19PM (#35923211)
I wonder if it would be better to scale down cluster paradigms (MPI stuff), rather than trying to somehow scale up shared-memory paradigms.
Re:U.S. Focuses on Invigorating Chiplets' to Stay Cuting- Ed... by JonChesterfield on Friday May 12, 2023 @06:44PM (#35923476)
Dropping cache coherency is a big lever for performance. That's definitely more annoying to program against than a magically coherent model though.
Re:U.S. Focuses on Invigorating Chiplets' to Stay Cuting- Ed... by bee_rider on Friday May 12, 2023 @07:50PM (#35923943)
“Big lever” is a nice expression, successfully expresses both the power and unwieldiness of a thing.