Posts by Jisi

1) Message boards : BOINC client : Leopard (Message 13829)
Posted 14 Nov 2007 by Jisi
Post:

I've not been able to reproduce the iTunes issue, even though I'm running on one of the original Intel iMacs (2 cores at 2GHz.) I'm not set up to do video iChat so I haven't tested that.


Unfortunately things haven't improved here.

You might try with a video podcast, as it happens quite fast then.
2) Message boards : BOINC client : Leopard (Message 13650)
Posted 7 Nov 2007 by Jisi
Post:
Hi.

I haven't read the whole thread (yet) but I've posted a problem with BOINC & Leopard on the einstein@home forum and was redirected to this forum.

Additional Info: Im using BOINC 5.10.28
I'll try to monitor this thread and give additional information if needed

----
I'll just copy my msg from the e@h forum

I've running BOINC with E@H under OS X 10.5 and noticed a problem. I set BOINC to run always, but use only 1 core (I'm on a 1st gen CoreDuo iMac here).

When BOINC and E&H runs sound playback in iTunes gets "jerky" e.g. every few seconds there's a hiccup (like some buffer is running empty). The problem vanishes when closing BOINC.

Also this is not Itunes specific, because it happens with VLC to (albeit it may be using the same QT libs iTunes does).

I already tried manually renice the e@h process to 19, but this doesn't help.

Another interesting thing is, that Leopard seems to ping-pong the process fomr one core to the other: Looking at the activity monitor with e@h running and nothing else that demands any CPU power, I see e@h consuming 100% CPU on one core, while the other is idle and a few seconds later core 2 is us 100% while the other idles.

Whether this is a problem I do not know, since I haven't checked with Tiger because I had no such problems there. Perhaps someone with Tiger can check the behaviour. Normally switching cores would of course be bad for caching reasons, but since the CoreDuo has this fancy shared cache I am not sure whether sharing actually means that both cores can use the same cache entries (would be no prblems then) or whether it's just some sort of dynmically allocating the cache to each core.

I tried to find a way to bind the e@h process to a specific core (I think I've seen a pbind or something commando on some unix box which can do this), but haven't found a way to do so.


Has anybody experienced the same problems and might even propose a solution?

The PROBLEM I want to solve is jerky audio playback while running BOINC.

The whole mumbojumbo about cores, caches and such where just my speculations what might be the problem, and in fact they may be totally irrelevant :)

Thanks
----




Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.