From: Leigh Orf <orf@scat.ssec.wisc.edu>
Newsgroups: comp.parallel.mpi
Subject: messing with CLOG files
Date: 24 Mar 1999 20:43:42 GMT
Organization: his wits' end
Message-Id: <7dbipu$qic$1@news.doit.wisc.edu>
User-Agent: tin/pre-1.4-971215 (UNIX) (Linux/2.2.3 (i686))
Xref: ukc comp.parallel.mpi:4827


I would like to write a BIG clog file, describing the state of my model
for a full run, which is on the order of 1000 timesteps. Is there a
way to split a clog file into individual timesteps, or timeblocks, so
jumpshot can handle them? Or, additionally, if there is a way to dump
small (only a few timesteps) clog files with different names on the fly
that would be great. I want to do this because I am trying to find a
sporadic bug that slows a node of my model down and I don't know its
location ahead of time, and I won't even try to feed jumpshot a 100 MB
clog file!

I've tried messing with MPE_Start_log, MPE_Stop_log, and MPE_Finish_log
in my code but get SIGSEVG: 11 (I'm not so sure what I'm doing here).

Also, are there programs other than jumpshot that read clog files?
Jumpshot is really slow on my Linux box since it's being interpreted.

Leigh Orf