by I'm new to the idea of real-time operations, and I notice
that Linux supports "soft" real-time via sched_setscheduler()
and "hard" real-time via the RT-Linux kernel patches.
I'd greatly appreciate if someone could discuss the tradeoffs among
using a kernel device driver, the RT-Linux extensions, and the
sched_setscheduler() calls for a given application. How would I know
what is a good match to an application? Or would applications use
several of those methods together? The examples I've seen for
RT-Linux sound like things that a device driver should handle. Why
not use device drivers instead? Or should the suggestion given in one
of the RT-Linux papers be followed: rewrite the time-dependent device
drivers as RT tasks?
Also, has anyone measured the maximum interrupt latency for RT-Linux?
Thanks for any insights.
-Rob
--
Robert W. Brewer