Steve Baker (sbaker++at++link.com)
Tue, 23 Dec 1997 11:02:42 -0600 (CST)
> Our application (architectural walk/flythrough) exposes memory bloat
> when being run over a long time. After a 15h test run (using gmemusage) it grew from
> overall 218MB to 378MB. I browsed the discussions about memory
> fragmentation in the mailing list archives and put the following
> statements into the code:
>
> amallopt(M_MXCHK, 1000000, pfGetSharedArena());
> amallopt(M_FREEHD, 1, pfGetSharedArena());
> amallopt(M_MXFAST, 64, pfGetSharedArena());
> amallopt(M_GRAIN, 64, pfGetSharedArena());
>
> Another 15h run then showed 70MB less memory usage, but that's still
> an 80MB plus. Using gmemusage, I found out that all of the bloat
> happens in the draw process, which I took as an indication of the
> fragmentation phenomenon.
Your experience directly mirrors ours. We spent a *lot* of time looking
for memory leaks - and found none - concluding as you did that this is
some kind of fragmentation. We didn't find an 'amallopt' strategy
that made much difference - but I'll try yours and see if it helps
us too. If it does then maybe the pfTeam should consider adding the
amallopt calls into Performer.
For what it's worth, we found that the amount of memory bloat per
hour tends to gradually level off until eventually there is no further
increase. However, this is an asymptotic behaviour and it can take
literally days of continuous running for that to happen. For us, the
final memory consumption was typically around twice the original amount
of memory consumed.
Anyone who suspects that they are suffering from this should try
plotting a graph of memory usage versus time. If it asymptotes out
then it's probably fragmentation - if it's linear then it's a
good old fashioned memory leak.
> Oh, the system is an Onyx2/IR2 with IRIX 6.4, Performer2.1, 1.5GB RAM.
We have seen this phenomenon on pretty much all platforms and all OS
varients. It's still a problem in Perf2.2 too.
Steve Baker 817-619-8776 (Vox/Vox-Mail)
Hughes Training Inc. 817-619-4028 (Fax)
2200 Arlington Downs Road SBaker++at++link.com (eMail)
Arlington, Texas. TX 76005-6171 SJBaker1++at++airmail.net (Personal eMail)
http://www.hti.com http://web2.airmail.net/sjbaker1 (personal)
** Beware of Geeks bearing GIF's. **
=======================================================================
List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
Submissions: info-performer++at++sgi.com
Admin. requests: info-performer-request++at++sgi.com
This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:56:28 PDT