nvidia chip along with skylake chip....

[Arto Pekkanen]

Stefan Wendler kirjoitti 05.03.2016 18:22:
> On 03/05/16 16:52, Arto Pekkanen wrote:
>> I am not sure if the main X-screen needs to be hardware accelerated or
>> not. Maybe, maybe not.
> That is the thing I do not know. But I was thinking about that myself.
> If the VESA driver reads from the VRAM there might be a chance that it
> will work. But then I do not know how the wiring of this
> IntelGPU-VRAM-nvidia dtuff is working. Maybe I will try it out. But 
> then
> Haswell support is pretty near and I have no problems waiting a little
> bit longer. If I need 3D, I boot into my Gentoo Installation.

VESA driver will not read from the nVidia VRAM. Neither does Intel 
driver.

The way it currently works is that VirtualGL intercepts function calls 
to libGL, and executes these libGL calls on separate X-server context, 
meanwhile forcing the primary GL surface to be rendered into a pixel 
buffer. During GLDraw function call VirtualGL will then copy the pixel 
buffer back onto the GL surface (with render to texture primitives) of 
the actual GL application.

In fact, this could work just fine even with the main X-server running 
on VESA as long as the libGL implementation is able to copy the pixel 
buffer data effectively enough into the software accelerated GL surface 
of the main X-server.

>> 
>> Anyway, you gotta remember that all the applications that you run with
>> VirtualGL (which dispatches GL stuff to the nVidia GPU running on a
>> secondary X.org) must be run inside a chroot where the libGL is the 
>> one
>> from the nVidia driver package. Otherwise things just won't work.
>> 


-- 
Arto Pekkanen