Quantcast

Mesa, Nvidia and libglvnd (the good, the bad and the ugly) :)

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Mesa, Nvidia and libglvnd (the good, the bad and the ugly) :)

arch dev mailing list
Nvidia provides for several releases libglvnd support which mean the hability
to have serveral libgl installed without hacks (nvidia-libgl, mesa-libgl
currently). With Fedora providing default Mesa with libglvnd support (with
some fedora patches, some upstream pending patches), i plan to provide Mesa
with libglvnd support with mesa-17.0

What does this means ?
Mesa and Nvidia will be able to live together in a perfect harmony :) Mesa and
Nvidia will not provide anymore mesa-libgl and nvidia-libgl packages; they are
useless. libglvnd will provide libgl support and will depend on an opengl-
driver, what Mesa and Nvidia will provide. Bumblebee will not work anymore
with Nvidia and Mesa, but only with prime (1).
Nvidia-340xx driver will still work with bumblebee but will need a specific
(non libglvnd?) Mesa version.
Nvidia-304xx driver will be (finally) moved to unsupported.

So, When ?
The first step is a new xorg-server version with upstream patches to improve
outputclass support and the glx.so symlink hack (2), now using ModulePath xorg
option (3) and adapt Nvidia and Mesa drivers. Let's hope i will not break
everything :)

The next step (after Mesa 17.0 release) is the libglvnd move.......

(1) https://devtalk.nvidia.com/default/topic/957814/linux/prime-and-prime-synchronization/
(2) https://bugs.archlinux.org/task/26284
(3) man xorg.conf

--
Laurent Carlier
http://www.archlinux.org

signature.asc (499 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Mesa, Nvidia and libglvnd (the good, the bad and the ugly) :)

Jerome Leclanche-2
On Sun, Feb 12, 2017 at 8:22 PM, Laurent Carlier via arch-dev-public
<[hidden email]> wrote:

> Nvidia provides for several releases libglvnd support which mean the hability
> to have serveral libgl installed without hacks (nvidia-libgl, mesa-libgl
> currently). With Fedora providing default Mesa with libglvnd support (with
> some fedora patches, some upstream pending patches), i plan to provide Mesa
> with libglvnd support with mesa-17.0
>
> What does this means ?
> Mesa and Nvidia will be able to live together in a perfect harmony :) Mesa and
> Nvidia will not provide anymore mesa-libgl and nvidia-libgl packages; they are
> useless. libglvnd will provide libgl support and will depend on an opengl-
> driver, what Mesa and Nvidia will provide. Bumblebee will not work anymore
> with Nvidia and Mesa, but only with prime (1).
> Nvidia-340xx driver will still work with bumblebee but will need a specific
> (non libglvnd?) Mesa version.
> Nvidia-304xx driver will be (finally) moved to unsupported.
>
> So, When ?
> The first step is a new xorg-server version with upstream patches to improve
> outputclass support and the glx.so symlink hack (2), now using ModulePath xorg
> option (3) and adapt Nvidia and Mesa drivers. Let's hope i will not break
> everything :)
>
> The next step (after Mesa 17.0 release) is the libglvnd move.......
>
> (1) https://devtalk.nvidia.com/default/topic/957814/linux/prime-and-prime-synchronization/
> (2) https://bugs.archlinux.org/task/26284
> (3) man xorg.conf
>
> --
> Laurent Carlier
> http://www.archlinux.org

Awesome!

J. Leclanche
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Mesa, Nvidia and libglvnd (the good, the bad and the ugly) :)

arch dev mailing list
In reply to this post by arch dev mailing list
Hi,

Thanks for thinking about libglvnd support in Arch. ;)

Le 12/02/2017 à 19:22, Laurent Carlier via arch-dev-public a écrit :

> Bumblebee will not work anymore with Nvidia and Mesa, but only with prime (1).

This sentence doesn’t really make sense. Bumblebee doesn’t work with
PRIME, they are two different solutions to Optimus support under Linux.
Some more correct statements:

– Bumblebee will still work with VirtualGL, but some changes will be
required for primus [1] (I’m not sure since I did not have time to test
it yet, but it could be as simple as patching bumblebee/primusrun for
exporting __GLVND_DISALLOW_PATCHING=1 when running optirun with primus
or primusrun, anyway I asked the person who fixed it in Fedora what
changes this has required);

– what will change with Nvidia PRIME support is that it would be easier
to switch from a X server running on Intel to a one running on Nvidia,
like nvidia-prime allows on Ubuntu, because you won’t have to switch
libgl packages on-the-fly. That’s a big win from an usability POV.

> Nvidia-340xx driver will still work with bumblebee but will need a specific
> (non libglvnd?) Mesa version.

Hum… What makes nvidia-340xx situation different from nvidia-304xx, and
without even talking of Bumblebee, does it even work with libglvnd?
AFAIK, the support was introduced in 361 series, and wasn’t backported
to the 340xx branch.

Regarding Bumblebee, I don’t think it would require anything specific,
but I should be able to test that and work around once everything will
have landed. ;)

Regards,
Bruno

[1] https://github.com/amonakov/primus/issues/193


signature.asc (531 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Mesa, Nvidia and libglvnd (the good, the bad and the ugly) :)

arch dev mailing list
Le dimanche 12 février 2017, 20:17:07 CET Bruno Pagani a écrit :

> Hi,
>
> Thanks for thinking about libglvnd support in Arch. ;)
>
> Le 12/02/2017 à 19:22, Laurent Carlier via arch-dev-public a écrit :
> > Bumblebee will not work anymore with Nvidia and Mesa, but only with prime
> > (1).
> This sentence doesn’t really make sense. Bumblebee doesn’t work with
> PRIME, they are two different solutions to Optimus support under Linux.
> Some more correct statements:
>
Doh, yes, really badly worded. I had the wild idea that bumblebee wasn't
supporting libglvnd, and i was wrong.

> – Bumblebee will still work with VirtualGL, but some changes will be
> required for primus [1] (I’m not sure since I did not have time to test
> it yet, but it could be as simple as patching bumblebee/primusrun for
> exporting __GLVND_DISALLOW_PATCHING=1 when running optirun with primus
> or primusrun, anyway I asked the person who fixed it in Fedora what
> changes this has required);
>
> – what will change with Nvidia PRIME support is that it would be easier
> to switch from a X server running on Intel to a one running on Nvidia,
> like nvidia-prime allows on Ubuntu, because you won’t have to switch
> libgl packages on-the-fly. That’s a big win from an usability POV.
>
> > Nvidia-340xx driver will still work with bumblebee but will need a
> > specific
> > (non libglvnd?) Mesa version.
>
> Hum… What makes nvidia-340xx situation different from nvidia-304xx, and
> without even talking of Bumblebee, does it even work with libglvnd?
> AFAIK, the support was introduced in 361 series, and wasn’t backported
> to the 340xx branch.
>
yes, nvidia-340xx doesn't provide any libglvnd support, and i don't know if
it's planned

> Regarding Bumblebee, I don’t think it would require anything specific,
> but I should be able to test that and work around once everything will
> have landed. ;)
>

Thanks Bruno for explaination about bumblebee situation
 
> Regards,
> Bruno
>
> [1] https://github.com/amonakov/primus/issues/193


--
Laurent Carlier
http://www.archlinux.org

signature.asc (499 bytes) Download Attachment
Loading...