performance analysis using spin

Hi,
we are currently using the tool “spin” (version 1.10) with QNX 6.2.1
on an x86 machine for a performance analysis. Unfortunately we get some unplausible results.
E.g. we get a CPU load of about 19000% when starting spin -c.
That seems to be a little bit too much, even though we know, that
our application can easily provoke a load of 100%. Is there a reason for
that enormous number. And possibly a way to reinterpret this result.

Thanks in advance
jba

SPIN is not an official QNX product. I would suggest you take a look at the source code …

See below discussion for SPIN source or you want to contact the author:

openqnx.com/PNphpBB2-viewtopic-t3247-.html

Not sure if it is included with 6.2.1, but you should search for ‘hogs’. It is more reliable than ‘spin’.

I made some adaptions to the 1.10 version of ‘spin’, so that ist also runs under QNX6.3
(specially memory and NIC statistics), if someone’s interested.

(Though: I didn’t manage to contact Igor yet, to post the changes to him, but hopefully this will happen within the next few days).

Yeah, I’m interested ! I started doing it a few weeks ago, but got pre-empted by something else, and then it slipped my mind…

Rennie

I made some adaptions to the 1.10 version of Sorry for the late answer, I do not attend to this NG regularily (in fact I hate
HTML-based newsgroups :slight_smile: )

I actually made a V1.11 of ‘spin’ wich works quite well actually, which I’d like to make available for everyone in OpenQNX (if Igor doesn’t oppose :slight_smile: ).

Does anybody know, how to do this?

Where don’t you attach it to your post so that everyone can “see” it.

Igor probably won’t oppose. As a matter of fact, he’s probably not even interested in QNX anymore.

Ok, here’s the V1.11 of spin.

All my changes are for Neutrino V6.3.0 and later only, so that this version should still run under QNX 6.2.

From the change note in spin.c:

Change log:
1.11 - Adaption to QNX6.3.0, June 8, 2005, by Karsten Hoffmann (kphoffmann@ish.de),
- New interface for NIC statistics again :frowning:
- Memory statistics interface changed as well.
- Date format changed to ISO-type

P.S.: Does anybody know, why this BBS complains about .tgz’ attachments, but it allows '.tar.gz’ files?! Wierd …

I find the QNX System Profiler the most reliable tool to measure performance, and I see who is running when.