I’ve been trying to figure out how to maximize CPU cycles available
to my SDL application (it’s a real time data collection and video
presentation system) on a dual-processor Linux box. Does
anyone know whether once the SDL video subsystem is init’d in DGA
mode (XFree86 >=4.1.0) the DGA commands still required the server
to be awake?
What I’m getting at is does SDL simply get direct access to the
framebuffer in DGA mode, meaning it can run 100% independent
of the server, or does it still has to communicate with the Xserver
to get things done.
If the former, then switching to fbcon or directFB drivers would buy
nothing, but if the latter, than the fbcon or directFB drivers would
give me tighter control on cpu and video synchronization by taking the
Xserver out of the scheduler queue (my application typically runs at
reniced to max-priority). Also I would be able to more accurately
predict the time required for particular SDL operations if I knew
it didn’t require going back the (possibly sleeping) Xserver…
Anyone got any thoughts on this??
office: 510-643-3573 mail: 3210 Tolman Hall #1650
lab: 510-642-1950 U.C. Berkeley
fax: 801-697-4179 Berkeley, CA 94720-1650