On Mon, 4 Oct 1999, Paul Koning wrote:
>
> Wiggins> While RTLinux is a very interesting and worthwhile project
> Wiggins> it didn't help too much :( Beyound clarifying a few things
> Wiggins> and bringing about a few minor idea's its the wrong things
> Wiggins> for what I'm looking for. Particularly the RT executive
> Wiggins> itself is not preemptable, the RT threads run in a single
> Wiggins> address space as well. These are the problems I'm looking to
> Wiggins> solve. I have some idea's but they still need
> Wiggins> formalising. Also no one seems to be interested in
> Wiggins> Microkernel development (due to asumptions about
> Wiggins> performance) which makes it hard to get feedback.
>
> You might look at RTAI (Paolo Montegazza); as I recall it uses those
> words. I haven't looked at it, thought. Don't know of any ARM
> porting effort for that one.
>
> As for being non-preemptable, in my mind that's a benefit, not a
> drawback. I know of real time systems that have failed due to relying
> on priority mechanisms and pre-emption. If you design your system
> without preemption and with a single priority level, you find yourself
> having to pay attention to guaranteeing that everything that must get
> done, will get done. That's the way you create reliable systems...
Priority is build into L4 it uses hard priority with round robin
on the same priority level. After thinking about the pre-emption issue
though I'm not sure its needed since the critical syscall IPC is about 1us
on a 100Mhz mips machine I hope to get something similar for the SA-1100.
a big problem though is address space context switches being expensive and
potentially interfering with RT tasks.
Maybe its about time to move this discustion to another list or
private or does no one mind it being here?
Cheers Adam
> > paul >
unsubscribe: body of `unsubscribe linux-arm' to [EMAIL PROTECTED]
