Re: [Dri-users] Using XF86 4.0 on Linux to remotely admin SGIs

[Randall H. <aa...@ip...>]

Georg Steffers:
 |I am not sure if it is possible to get accelerated 3D-Graphics remotely
 |but i dont think so...for hardware acceleration of 3D i think you need
 |direct access to the Hardware the OGL lib is written for and not only
 |communication over the X-Protocol (only a guess, perhaps someone can
 |correct me...).  You should use a card that is good supported by XFree
 |(look on the XFree86 Webpages) it may not make a lot of difference how
 |fast your graphics Hardware is, because your bottleneck is the Network...
 |I hope that helps you out....

I believe it depends on how fast your net and graphics systems are.  I run
3D OpenGL apps from an SGI Onyx over the Internet (via ssh) to a FreeBSD
box equipped with cable modem using GLX with significant speedup
(displaylist'ing the core rendering loops improves performance even more).
My GLX is hardware-accelerated.  I'm not using DRI yet however (G200/G400
needs more work).

In general, from Precision Insight's DRI architecture:

    http://www.precisioninsight.com/dr/mini_poster.jpg

it doesn't appear there's anything that prevents you from getting hardware
acceleration with the new system when pushing OpenGL remotely via GLX.
True, the OpenGL commands will have to pass through the X server (and
kernel potentially), but it appears the X server and kernel module can in
general be hardware-accelerated (i.e. remote apps aren't necessarily forced
to use "Mesa SW Renderer").

Someone please step in if you have more details.  I haven't had a chance to
dig into this much yet.

Randall