Funny WU


log in

Advanced search

Message boards : Number crunching : Funny WU

Author Message
Mark Lybeck
Send message
Joined: 9 Aug 99
Posts: 209
Credit: 103,364,900
RAC: 80,616
Finland
Message 1352271 - Posted: 30 Mar 2013, 23:04:14 UTC

Check this WU it seems that no-one can make sense of it....

http://setiathome.berkeley.edu/workunit.php?wuid=1196302413


____________

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8633
Credit: 51,545,205
RAC: 48,352
United Kingdom
Message 1352274 - Posted: 30 Mar 2013, 23:10:34 UTC - in response to Message 1352271.

It's a perfectly normal MAX_TRIPLETS_ABOVE_THRESHOLD (Exit status -12)

Check this WU it seems that no-one can make sense of it....

http://setiathome.berkeley.edu/workunit.php?wuid=1196302413

And there's a button (above right when posting) to make urls clickable.

Profile MikeProject donor
Volunteer tester
Avatar
Send message
Joined: 17 Feb 01
Posts: 24524
Credit: 33,855,373
RAC: 23,795
Germany
Message 1352275 - Posted: 30 Mar 2013, 23:12:33 UTC

Thats a normal work unit with to much triplets.
All cuda hosts failed while the CPU hosts didn`t.

____________

Profile Michel448a
Volunteer tester
Avatar
Send message
Joined: 27 Oct 00
Posts: 1201
Credit: 2,891,635
RAC: 0
Canada
Message 1352316 - Posted: 31 Mar 2013, 4:49:28 UTC - in response to Message 1352275.

Thats a normal work unit with to much triplets.
All cuda hosts failed while the CPU hosts didn`t.



triplets ?

Spike count: 0
Pulse count: 31
Triplet count: 0
Gaussian count: 0


oh ya, 0 is too many
____________

Josef W. SegurProject donor
Volunteer developer
Volunteer tester
Send message
Joined: 30 Oct 99
Posts: 4302
Credit: 1,072,504
RAC: 1,211
United States
Message 1352343 - Posted: 31 Mar 2013, 7:15:27 UTC - in response to Message 1352316.

Thats a normal work unit with to much triplets.
All cuda hosts failed while the CPU hosts didn`t.



triplets ?

Spike count: 0
Pulse count: 31
Triplet count: 0
Gaussian count: 0


oh ya, 0 is too many

It's the "cudaAcc_find_triplets doesn't support more than MAX_TRIPLETS_ABOVE_THRESHOLD numBinsAboveThreshold in find_triplets_kernel" version of the -12 error. The analysis requires keeping track of which bins are above threshold, and for CUDA there's limited storage considering that many arrays are being analyzed in parallel. Anyhow, just because there are more bins above threshold than normal does not mean that the particular spacings between those bins will be correct for detecting a Triplet event.

When we make the transition to SETI@home v7 those errors will no longer happen. Jason coded the new CUDA apps to send those rare cases back to be rechecked on the CPU. But meanwhile those -12 errors are a fact of life for stock CUDA SETI@home Enhanced applications.
Joe

Message boards : Number crunching : Funny WU

Copyright © 2014 University of California