Usualy, real time driver should not use big lock or take the bus for a long time. If the video driver send a burst of 10 ms length on the PCI bus, you could not have a system that react faster that use the bus. I have heard complain about NVIDIA driver that does such thing.
The driver should also be design with maximum time for each system call from an application (like 1ms max) like any real time system. When the first RT patch was introduced in linux, they mesure some code for page managment with some delay more than 300 ms. So driver for real time application need to be fast to keep the cpu or buses for himself as short as possible. Not using some general lock enable kernel preemption, that could reduice latency for application that need the cpu. I don't know if linux use an IO scheduler for there 2006/8/29, Lourens Veen <[EMAIL PROTECTED]>:
How are we on hogging the PCI bus? Are the low-latency Linux people (you know, the ones who build digital audio workstations and hard disc recorders and hack on Ardour and Jack and LADSPA) going to complain about us messing up the latency to their PCI sound cards? I just happened to come across the Ardour web site, who have some nice bashing of VIA chipsets and a lot of praise for Matrox cards on their hardware page. Obviously, quality hardware will support this kind of use, but I wonder if it's been taken into account explicitly and where we are on this currently. Lourens _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
_______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
