Re: ADA os talk

[Robert Spooner <rls19@psu.edu>]

Larry Kilgallen wrote:
> 
> In article <9mgg0p$26j$1@nh.pace.co.uk>, "Marin David Condic" <dont.bother.mcondic.auntie.spam@[acm.org> writes:
> 
> > Like I said, my memory of the situation whas that techniques were possible
> > under VMS to run realtime, but that VMS wasn't designed to specifically
> > provide some kind of realtime mode. Sort of a case where if you had enough
> > privileges and enough priority and could keep anything else from taking
> > precidence, you could convince yourself that you had fixed latency and
> > weren't going to get swapped out. This is different than having a process
> > that the OS manages in realtime mode and guarantees you that none of this is
> > going to fail.
> 
> Absolutely VMS is designed so that you need privilege to alter its
> behavior.  By default there is fair treatment for all, and that is
> certainly inadequate for realtime needs.  The controls to tweak,
> however, are all documented as being "for realtime", and those of
> us who pass them every day (and ignore them) tend to presume the
> priority mechanisms are adequate.  The scheduling algorithm has
> a different behavior in the upper half of the priority range,
> for instance, specifically for realtime.

VMS is useful for a wide variety of things, including some near
real-time applications, but the predictability of how long it will take
for the O/S to perform certain operations was not good enough for us,
and it also takes too long to boot.  We ended up using VAXELN for our
real time O/S on the KAV30, a real-time VME single-board VAX, and VAXELN
Ada for our applications.

VAXELN has more deterministic algorithms and is an execution only
environment with address space protection and a lot of other nice things
that some other real-time O/Ss are just now coming out with. 
Unfortunately, after years of DEC preaching "we have the migration path
you need" they did not port VAXELN to the Alpha, did not move to Ada 95,
"retired" VAXELN Ada and DEC^H^H^HCompaq is in the process of retiring
the rest of the environment we were using.  It's really sad, because the
quality of the hardware, software, and support was in a class by itself.

We were able to use the same compiled code and file system (except for
O/S dependent interfaces in a few package bodies) on both VMS and
VAXELN.  Try that with a Sun/Solaris host and 68060/VxWorks target.  At
least we have GNAT for both those environments.  Even the default record
layout was the same since we used the VAX processor for host (data
reduction) and target (experiment) so we didn't have to use a lot of
record representation clauses.  It was as seamless as could be.  We had
the advantages of both a real-time operating system and a good host O/S
with that setup.

Bob
-- 
                            Robert L. Spooner
                     Registered Professional Engineer
                       Associate Research Engineer
                  Intelligent Control Systems Department

         Applied Research Laboratory        Phone: (814) 863-4120
         The Pennsylvania State University  FAX:   (814) 863-7841
         P. O. Box 30
         State College, PA 16804-0030       rls19@psu.edu