Re: Two computers in one: two users each with their own accounts, monitor, and keyboard?
Boyd Stephen Smith Jr. put forth on 1/12/2010 11:41 AM:
> On Monday 11 January 2010 23:45:12 Stan Hoeppner wrote:
>> Boyd Stephen Smith Jr. put forth on 1/11/2010 12:23 PM:
>>> 4. I had to switch VTs to the X server that was handling the OpenGL
>>> commands for the GLX calls to complete. Likely, the video driver I am
>>> using requires exclusive access to the hardware to process some GLX
>>> requests.
>>
>> This is my point. The Linux OpenGL architecture has been optimized for
>> local hardware rendering. It's impossible to get "usable" remote OpenGL
>> today, mainly because textures are such an integral component of 3D
>> rendering today. Trying to push texture data over ethernet (even GigE)
>> for real time rendering is not really doable.
>
> That's not due to any Linux (kernel) limitation that "multiuser 3D OpenGL
> would no longer be supported" and a far cry from having "to eliminate over the
> network OpenGL completely".[1]
>
> It takes time to push data across the network -- always has. In fact, network
> speed on (on average) faster today than when SGI or SUN workstations were big
> sellers. While an apples-to-apples comparison is probably impossible, I'd
> wager that using GLX on a remote X session is actually faster now than "back
> in '02 or '03" or even the SGI workstations.
>
> Multi-pass rendering of a scene with 1G of textures and several megs of
> compiled shader programs in 1/60th of a second is tough, but if your X server
> hardware is good enough, your bands wide enough, and your round-trip times low
> enough, standard Xorg will do it without missing a beat.
>
> [1] If you don't recognize the quotes, they are from your earlier messages in
> this thread.
I recognize it. As a test of your theory, fire up your favorite ID Software
OpenGL Linux game on one Debian PC piping its GL calls over the network to
another PC's X server. Run the game at 640x480x16bpp windowed. Le me know how
this works, or doesn't, and what errors you generate. Also report how many
hours/days you spend troubleshooting it to make it work. ;)
--
Stan
Reply to: