From: Prakash M. <pma...@da...> - 2005-04-24 08:17:27
|
> > I think it is difficult to make a micro-kernel realtime also. The most > that > > can be achieved is "soft" realtime. Microkernels are best with > asynchronous > > system calls, and realtime means all things have to synchronize > perfectly. > > It could be a little contradictory, but perhaps there is a way... > > > > To my understanding microkernel design is not a constrant for being > realtime. Our main focus should be to minimise latencies(mainly > scheduling latency and interrupt latency) as much as possible. Infact > Most Popular Realtime OS are Microkernel. One being the Pioneer in > RTOS QNX(www.qnx.com). I was inspired by their design. By making > realtime kernels we would neither be supporting 8051's and variants > nor PICs (it would be very difficult or almost impossible). We would > be supporting high end embedded chips like ARM(they are one of the > leader of embedded world market after Intel released StrongARM), > PowerPC, x86 etc. Message passing overhead is not much issue in these > systems. I just went thru the QNX website and found it to be too great to be true! Oops, since it is there, it must be true. How about we model manrix after QNX? Are there any copyright issues involved? I don't think so. Maybe some algorithms are proprietary but the concept is up there for grabs, I guess and they obviously don't reveal the real trade secrets on the web... Does anybody know anyone in Kathmandu or nearby who has extensive experience with the QNX? If yes, we could consult with him/her. Otherwise, no big fuss; we can carry on! Regards, Prakash. |