Subject: | |
From: | |
Reply To: | |
Date: | Thu, 13 Jun 1996 17:20:04 GMT |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
[log in to unmask] (ChipDorm) wrote:
>I have been observing our HP3000/980 through Glance
>and have noticed that it is switching to Compatibility
>Mode alot (>1800/sec). Searching the directory, I have
>found only one PROG in use with the rest NMPRG. I assume
>that we have recompiled all of our programs and libs so I wonder
>what is causing the high number of CM switches. Incidentally,
>I see that most of our KSAM files are the MPE-type (KSAM/KSAMK)
>and wonder if MPE/XL switches to CM to access these files.
>
>Does anyone have a list of the "usual suspects" that
>cause high CM switches under MPE/XL?
>
Hi Chip-
You are right; >1800/sec is too high and you have a lot to gain by
fixing this. There is a long list of "usual suspects" though, and
they differ from release to release. At a minimum, KSAM/KSAMK files
are definitely CM and I would recommend using FCOPY to convert these
guys to NM KSAM. Prior to MPE/iX 5.0, Message files are in CM. If
you use a lot of these, the best solution would be to go to 5.0. (I
think that is a good idea for everyone anyway.) Other CM files
include RIO and Circular files.
I think that if you are doing sorts that do anything "out of the
ordinary" such as alternate collating sequence, then sort still
switches back to CM, though I haven't checked recently.
If after doing all of this you still can't account for some of the
switches, you could put a system wide breakpoint in the switch routine
that can log a stacktrace and then continue. But get some expert
advice before doing this. At a minimum, performance will go to hell
while you are doing this. At a maximum, well....let's not talk about
that.
Hope this helps,
/Steve
|
|
|