Hi Tom, I'm developing quite a similar system for experiments in our Oculomotor Lab.
The graphics framework does already work quite well. There are only some rare occurrences where the SCHED_FIFOed graphics thread hasn't finished drawing the next frame on VSYNC. I guess they are due to some kind of priority inversion in thread synchronisation. Another try would be with the low latency patches as we are still using a stock 2.6 kernel. CPU and system load is really low (~ 5% with jumping point paradigm). I haven't looked further into these timing issues however since I'm just occupied with implementing a Comedi link for data acquisition and experiment control. My current development setup: - AMD Duron 800 system - Matrox G550 - Debian Sarge with kernel 2.6.8 - CVS versions of DirectFB, DFB++, DirectFBGL - Villes Mesa-embedded-2-branch-2004-07-15 with freeglut-directfb - Comedi 0.7.69 Regards Jan
