[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kwin
Subject:    Re: Supporting Hybrid Graphics
From:       Fredrik =?iso-8859-1?q?H=F6glund?= <fredrik () kde ! org>
Date:       2013-07-17 21:42:17
Message-ID: 201307172342.17121.fredrik () kde ! org
[Download RAW message or body]

On Monday 15 July 2013, davyaxel@free.fr wrote:
> Hello guys,
> 
> I'm looking for implementing Hybrid Graphics support for KWin.
> 
> The goal is to have an UI option 'use XXX graphic card' that will make \
> the application to use the XXX graphic card. 
> Today there are two way for users to use hybrid graphics:
> 
> .On Intel/Nvidia, Intel/Amd open source driver, the application to render \
> on the discrete graphic card is launched with a path indicating the \
> discrete card driver and on a second server X that will be redirected via \
> VirtualGL. 
> .On Intel/Amd catalyst, We can switch with the Amd driver which card we \
> want to use and the whole desktop will use this card (after an X \
> restart). 
> On Windows and muxless(recent) Intel/Amd, Amd catalyst let the user \
> choose which graphic card use applications (there are default settings we \
> set of course). I think we could support a similar system by an UI \
> option. 
 
This information is outdated. The solutions you are talking about have been
replaced by DRM PRIME and provider objects in Xrandr 1.4.

XRRSetProviderOutputSource() and XRRSetProviderOffloadSink() are used
to configure a provider to act as an output source for another provider,
or as a render offload device.

The DRI_PRIME=<provider> environment variable determines which GPU
is used by individual applications.

This is supported by the open source drivers and the NVIDIA driver, \
although the NVIDIA driver can only act as an output source provider in the \
current version.

Note that GPU offloading only works when compositing is active.

Regards,
Fredrik

_______________________________________________
kwin mailing list
kwin@kde.org
https://mail.kde.org/mailman/listinfo/kwin


[prev in list] [next in list] [prev in thread] [next in thread]