What are w crunching?
15 Feb 2012 10:13:26 UTC
Topic 84824
(moderation:
My pc's crunch for Albert since a couple of weeks now. I've seen apps failing and they were improved; I haven't got errors or invalid results now for at least two weeks.
The apps and the project seem to be stable. That raises a new question: Do we crunch real (new) data or do we just recrunch already known data?
If we do just recrunching it makes more sense to give Einstein more resource share until a new app or a new version is there to be testet.
Can the project manage to send a notice if new apps are there for testing?
What are w crunching?
)
Hello Alex,
what we crunch you can see in the names which the programs have.
S6LV1 means, we crunch data from search six, which are also being crunched over at Einstein.
But we are testing a different search method. Called LineVeto, hence the 'LV' in the name.
So yes, we crunch 'old' data. But we do it in a totaly different way which should turn up different results for better conclusions.
Somebody will show up how can explain this better.
As for that you didn't have anymore invalids since about two weeks. That's nice and promising, but others still have them. If that percentage is still too high in the eyes of the developers, they will continue to improve the code.
I did just check your computers, the one on top of the list does have one invalid with OpenCL 1.21. And if you check all your other tasks you will find some where it is valid for You, but which was sent to three PCs at all because for one it was invalid.
Hope this helps,
Christoph
EDIT
One more thing, when the actual tested apps are finaly released, the devs may start improving the speed of them.
So there will be a new set which have speed improvements which needs testing again.
So it is basically never finished. Mybe some time till there are new apps out.
That will be published over the frontpage news, which will come to you over the News tab in BOINC Manager.
Christoph
The focus on Albert@Home is
)
The focus on Albert@Home is testing new server code (scheduler, web code) in preparation to use it on Einstein@Home.
However as the current server code doesn't allow to run more than one "locality scheduling application" (like S6Bucket or S6LV1) at a time, we have been testing S6LV1 over here, too.
S6LV1 has, by now been moved over to Einstein@Home, and consequently the workunit generator has been stopped here. We leave e.g. the validator running, though, such that people will still get credit for the tasks already in progress.
The BRP4 tasks sent here are copies of tasks that run over at Einstein@Home, too.
Here these are solely for testing the new OpenCL application and don't serve any other (scientific) purpose.
BM
BM
I have one BRP4SSE unit
)
I have one BRP4SSE unit running on Einstein and one BRP3SSE running on Albert. This is much slower on my Linux box, no GPU.
Tullio