Michael Margolin
Michael Margolin
@HassanKhadour Thanks for taking a look! I've updated the commit message to better explain the change and its use case.
@galpress @rajachan Agree it's about time for refactoring HW accelerators memory usage code. I'll take this on.
I've added abstraction for various memory types as well as support for Habana Labs AI accelerators. @galpress @rajachan @HassanKhadour I would appreciate your review.
@HassanKhadour Can you please share failing check results?
@HassanKhadour Can you please merge this PR or suggest some extra improvements? Thanks!
@rleon Our initial intention was making efa_setup_qp() more robust as its current implementation has internal calculations that, although very unlikely, can be tackled by external (seemingly valid) inputs. Agree with...
@galpress @rajachan @HassanKhadour any comments will be appreciated.
How is that this PR is still waiting for review after we've been asked to refactor perftest accelerators memory support and at the same time commits are pushed directly to...
> I merely suggested it doesn't make sense that the same code is copy-pasted again and again. Since you're already pushing other code that does it, might as well merge...