[Intel-gfx] [RFC 0/8] Force preemption

Chris Wilson chris at chris-wilson.co.uk
Fri Mar 16 22:34:09 UTC 2018


Quoting jeff.mcgee at intel.com (2018-03-16 18:30:57)
> From: Jeff McGee <jeff.mcgee at intel.com>
> 
> Force preemption uses engine reset to enforce a limit on the time
> that a request targeted for preemption can block. This feature is
> a requirement in automotive systems where the GPU may be shared by
> clients of critically high priority and clients of low priority that
> may not have been curated to be preemption friendly. There may be
> more general applications of this feature. I'm sharing as an RFC to
> stimulate that discussion and also to get any technical feedback
> that I can before submitting to the product kernel that needs this.
> I have developed the patches for ease of rebase, given that this is
> for the moment considered a non-upstreamable feature. It would be
> possible to refactor hangcheck to fully incorporate force preemption
> as another tier of patience (or impatience) with the running request.

So one big problem I should have realised earlier is ksoftirqd screws up
any latency sensitive operation. I have an ugly idea for a hack, but we
really should cap the latency problems with tasklets.
-Chris


More information about the Intel-gfx mailing list