yodaiken@chelm.cs.nmt.edu wrote:
> The larger POSIX standard seems basically pointless for hard RT, although
> I'm looking for ideas on how to allow for "soft rt" linux processes.
> One idea would be to use the RTL tasks to control scheduling of Linux
> processes in response to queue lengths and other hints. Suggestions?
If you're talking about RTL forcing a Linux-user-space process to "run
next", then YES, Victor, this is VERY important for creating 'modular
software architectures': where part of a job is done by a process in
Hard-RT, part by a Soft-RT process, and part by a processs relying on
the linux scheduler !
I would love it if I could "guarantee" that my soft-real-time
linux-user-space process will run 1-2 linux time slices after my 60 Hz
RT-periodic task (my HZ=1000 and a time slice is a millisecond).
This would rectify one of the drawbacks of RT Linux: anyt process that
must meet even a "soft" scheduling deadline must go into RTlinux space,
preventing it from using any plain-linux services (device drivers,
shared libraries, IPC, etc...). Moving processes into RTlinux space also
inhibits object oriented design (forget C++), and forget about using a
symbolic debugger.
Your idea would remove a serious RTL limitation, potentially winning
over those who use commercial Real Time OSes (which don't limit the
runtime environment as RTLinux currently does).
--
Steve Rosenbluth
Jim Henson's Creature Shop
2821 Burton St, Burbank CA
(818) 953-3030