In a message on 08-25-94, Louis Rizzuto said to Peter Hansen:

LR>In the past, on mainframes, especially for IBM, mainframes,
LR>multi-taking os's had the capability to define a high priority process
LR>(i.e. a task) that the os would give to that high priority task - for
LR>"real-time" tasks.  The general title to these specially defined high
LR>priority tasks were referred to as TP (telcommunication Processes). 
  :
OS/2 features four classes of task prioritization of the kind you
describe, with 31 levels of priority within each group.  The four classes
are: time critical, server, regular, and idle.  The time critical class is
similar to the TP stuff you mention, and would be used for similar
purposes.  Regular is for normal user programs, server is for server
programs (to make sure they get superior response over any regular
programs that happen also to be running on a server machine), and idle is
for non-critical daemons and monitors.

All this stuff is described in wonderful detail in "The Design of OS/2" by
H.M. Deitel and M.S. Kogan (ISBN 0-201-54889-5, Addison-Wesley, 1992). 
This is "required" reading for anyone doing OS/2 work at the level you
contemplate.

LR>I presume that somewhere in IBM's OS/2 literature - and perhaps in the
LR>C++ literature "real-time" is addressed - where might that
be for both
LR>OS/2 and C++, Peter? 
  :
C++: don't know.  There is really no direct connection between real-time
and C++ (as I and others have suggested, you can use pretty much any
language for real-time work, depending on your timing requirements.) 
Basically, the use of C++ has negligible impact on the performance of your
real-time system since the only performance overhead incurred is when
using C++-specific features such as virtual functions and object
construction/destruction.  None of the C++ features cause response/
run-time to become indeterminate, just a little longer.

As for OS/2 and real-time...hard to say.  I have scanned the indices of
many OS/2 books and rarely found entries for interrupt latency or response
time, let alone "real-time" or any variation thereof.  Even the
"Design of
OS/2" book mentions real-time only in the context of Ethernet (which
isn't) and Multimedia (which is, but on the level of the user, not process
control).  In fact, to date, every "official" mention of real-time in
relation to OS/2 has been some sort of "soft" real-time and not the hard
stuff relevant to most industrial control (excepting only things like
chemical processes that need only a few cycles a second or minute).

IBM has perhaps only a couple of employees (in the world!) who know much
of substance about real-time uses of OS/2.  Steve Mastrianni, an
independent consultant who "wrote the book" on OS/2 device drivers, was
hired (part-time?) by IBM as a consultant on real-time issues, probably
because he knows more than anybody within IBM.  This area is hard slogging
and you certainly can't easily find freely available examples of relevant
code (if any exists), for example.  Any progress you report in the area
would be very well received by (at least) me and probably many others.

LR>Would you have any contact information for the company that produces
LR>VXWorks, Peter?  Thanks much for your help and sharing your experience
LR>on these matters.
  :
I'll look into it, although I've only seen VxWorks mentioned in an ad in
some magazine (don't know which).  I suggest you ask in the FIDO Embedded
conference since most of the relevant magazines are read by at least one
participant or another and somebody else will be able to help.  If I come
across the info, I'll post there since it's not OS/2-related.

Cheers,
Peter Hansen  ***  Engenuity Corporation  ***  Guelph, Ontario, Canada
Internet: peter.hansen{at}canrem.com    RelayNet:->CRS    FIDO:(1:229/15)
___
 * MR/2 * ***  Learn Esperanto, the international language.  ***
--- FidoPCB v1.5 beta-'e'