[RFC xserver v3 00/11] DRI3 v1.2: DMA fences

[Nicolai Hähnle]

Will this also support the wait-fence part of PresentPixmap?

Having an explicit wait-fence in PresentPixmap could potentially help 
with OpenGL driver multi-threading. We currently have to sync with the 
driver thread before sending PresentPixmap, to ensure that the necessary 
command submissions have completed in the kernel and an implicit fence 
added to the buffer.

It would be better if we could just create a syncobj in the 
application/API thread, add that as a wait-fence for the present, and 
hand it off to the driver thread to be signaled once previous GL 
commands have completed.

Cheers,
Nicolai


On 06.11.2017 22:42, Louis-Francis Ratté-Boulianne wrote:
> Hello,
> 
> This patchset implements of what we would like to see become DRI3 v1.2,
> that is the implementation of DMA fences.
> For some context, please see previous submissions:
> 
> https://lists.x.org/archives/xorg-devel/2017-August/054439.html
> https://lists.x.org/archives/xorg-devel/2017-September/054770.html
> 
> The main change in this iteration is:
> 
>   - Make SetTriggered and Reset return a result instead of adding
>     a new 'type' field for fences.
> 
> Here are the repositories:
> 
> https://gitlab.collabora.com/lfrb/dri3proto/commits/rfc/2017-10/x11-fences
> https://gitlab.collabora.com/lfrb/presentproto/commits/rfc/2017-11/x11-fences
> https://gitlab.collabora.com/lfrb/xserver/commits/rfc/2017-11/x11-fences
> 
> Thanks,
> 
> Louis-Francis
> 
> _______________________________________________
> xorg-devel at lists.x.org: X.Org development
> Archives: http://lists.x.org/archives/xorg-devel
> Info: https://lists.x.org/mailman/listinfo/xorg-devel
> 


-- 
Lerne, wie die Welt wirklich ist,
Aber vergiss niemals, wie sie sein sollte.