Well, Gavin,

At first I was miffed at your need-to-know-basis response, but after
reviewing the documentation, it is clear why you responded as you did, that
is, it is highly dangerous ;-|

I'll have to have a little talk with the individual that thought he was
giving me a good lesson on MPE/3K internals, re: polling vs interrupt,
etc.... then I'll go read some manuals.

Have a great rest of the day.

-dtd

> -----Original Message-----
> From: Gavin Scott [mailto:[log in to unmask]]
> Sent: Monday, January 08, 2001 11:31 AM
> To: [log in to unmask]
> Subject: Re: nscontrol question
>
>
> Dave replies:
> > 1) How does one (or where does one find info to) mark a program
> > as resident?
>
> If you have to ask then you don't need to do it :-)
>
> > 2) Is it correct that MPE or HP3000s do not make good
> real-time machines
> > because all serial connections are polled rather than
> hardware interrupt
> > driven?
>
> Serial connections are not polled.
>
> > Perhaps the designers had imagined some specialty I/O cards that
> > would handle hardware interrupts?  IIRC someone once told
> me that neither
> > the OS nor the hardware were set up to handle hardware
> interrupts (except
> > maybe internal hardware, possibly the IEEE488, er... I mean HPIB).
>
> Completely untrue.  Both Classic and PA-RISC 3000s are fully interrupt
> driven.
>
> Having a real-time OS basically means being able to guarantee
> worst-case
> latencies on interrupt handling.  The problem in doing this
> is not that it's
> hard to create a process which will have a high enough
> priority, but that
> you have to guarantee that other interrupt-driven processes
> will never keep
> interrupts disabled longer than your worst-case latency requirement.
>
> For small values of "real" and large values of "time" MPE can
> be made to
> look like a reasonably real-time OS in some cases.  The
> question is not how
> to make MPE real-time, it's what do you want to accomplish,
> and what's the
> best way of doing so.
>
> G.
>