Quoting Chris Wilson (2018-01-22 15:41:06)
> If we remember to cancel the signaler on a request when retiring it
> (after we know that the request has been signaled), we do not need to
> carry an additional request in the signaler itself. This prevents an
> issue whereby the signaler threads may be delayed and hold on to
> thousands of request references, causing severe memory fragmentation and
> premature oom (most noticeable on 32b snb due to the limited GFP_KERNEL
> and frequent use of inter-engine fences).
> 
> Signed-off-by: Chris Wilson <[email protected]>
> Cc: Tvrtko Ursulin <[email protected]>

Poke. I've been staring at this problem since December, it would be nice
to get some resolution :)
-Chris
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Reply via email to