HP3000-L Archives

May 2002, Week 2

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Wirt Atmar <[log in to unmask]>
Reply To:
Date:
Wed, 8 May 2002 14:10:56 EDT
Content-Type:
text/plain
Parts/Attachments:
text/plain (40 lines)
Christopher writes:

> We contacted INLEX to see if they knew
>  how we could convert these into native mode programs. They pointed us to
>  the OCTCOMP.PUB.SYS file. We ran this on a couple of our larger programs
>  and have not noticed a difference in the amount of time these programs run.
>  We were informed that it should increase dramatically as soon as we run
>  them through the OCTCOMP program.

"Increase dramatically" may be too dramatic a phrase. About the best you will
experience after OCTCOMPing a program will be ca. 30% increase in
performance, which means that you'll have to be looking for it if you want to
measure it. A 30% increase or decrease in performance is well within the
normal variance that any program experiences during each of its runs and will
often be swamped on a busy machine by what else the machine is being asked to
do each time that you run your program.

That's not to say that it's not worth doing. You want to take your 30%
performance increases anywhere that you can find them. However many CPU cyles
it took to execute your program under CM, by OCTCOMPing it, you've freed up
30% of those cycles so that they can be used by some other process. That's
not a bad deal for typing just a line or two.

The primary reason that you won't notice that much improvement in converting
your CM program to essentially full-native mode through OCTCOMP is because
the run-time translator that instantly converts each CM instruction to its NM
equivalents is not that all that bad by itself, thus there's only a certain
percentage gain to be had by OCTCOMPing.

Nonetheless, I'm frugal ("cheapskate") enough to suggest that any CM program
that is run often should be OCTCOMPed to near-NM. As Stan is likely to
suggest, there are corner cases where OCTCOMPing a program may actually cause
it to run slower than the original CM program, but they're so rare as to be
not considered.

Wirt Atmar

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *

ATOM RSS1 RSS2