The City of Largo is going to deploy 500 thin clients with 3D video cards (ATI) in early 2007. We are testing Beryl and Fedora Core and it’s working pretty well over remote display. When started, it’s producing this message on the command line:
libGL error: open DRM failed (Operation not permitted)
libGL error: reverting to (slow) indirect rendering
Research indicates this is because of remote display. Beryl is running on the server, and pushing the image down to the thin clients.
Any enhancements to help with performance is greatly appreciated. The speed right now is about 80% of what is required for me to deploy. Thanks.
The D in DRM means “direct”, which means directly accessing the GPU, bypassing the network. That of course works only if there is a direct connection between CPU and GPU.
The warning is there because there is only a network link between the CPU (on the server) and the GPU (on the client). You can’t expect performance to be the same, all drawing commands have to be transmitted over the network.
To improve performance, you could try display lists. They are only uploaded once, so they should dramatically reduce network traffic. VBOs should also help.
Latest Nvidia drivers support some extensions that are required by the compositing manager and thus allow hardware-accelerated desctop effects. As far as I know ATI doesn’t support this currently.
Check that link… it’s not exactly what you are looking for, but…
“VirtualGL is an open source program which redirects the 3D rendering commands from Unix and Linux OpenGL applications to 3D accelerator hardware in a dedicated server and displays the rendered output interactively to a thin client located elsewhere on the network.”
And other point is that nvidia linux proprietary drivers do support (remote) accelerated indirect rendering. I think ATI and DRI/DRM don’t support this… but I’m not completely sure.
