N-Body simulation may be faulty over 800 hours computation time estimate
log in

Advanced search

Message boards : Application Code Discussion : N-Body simulation may be faulty over 800 hours computation time estimate

Author Message
Profile Marcin
Avatar
Send message
Joined: 18 Jun 09
Posts: 4
Credit: 1,845,212
RAC: 0

Message 58413 - Posted: 26 May 2013, 14:04:55 UTC

i'd like to report a serious bug with some strand of N-body simulation tasks:
some of them, namely all with "chisq_alt_" in their name show computation time estimate of over 800 hours, and get to about 10% after running 3 days straight, verification then shows as inconclusive on the website
here is example of 1 task
MilkyWay@Home N-Body Simulation 1.08
de_nbody_10k_chisq_alt_40913_1366886102_673146
computation size 624753000 GFLOP’s
it's estimated computation time was 799 hours, it seemed to be not progressing at all
i think there may be a bug related to that task series.
To milkyWay@Home team: please consider this as a bug report and please investigate the cause of this

Thanks in advance

mikey
Avatar
Send message
Joined: 8 May 09
Posts: 2038
Credit: 181,658,115
RAC: 274,185

Message 58418 - Posted: 27 May 2013, 11:39:16 UTC - in response to Message 58413.

i'd like to report a serious bug with some strand of N-body simulation tasks:
some of them, namely all with "chisq_alt_" in their name show computation time estimate of over 800 hours, and get to about 10% after running 3 days straight, verification then shows as inconclusive on the website
here is example of 1 task
MilkyWay@Home N-Body Simulation 1.08
de_nbody_10k_chisq_alt_40913_1366886102_673146
computation size 624753000 GFLOP’s
it's estimated computation time was 799 hours, it seemed to be not progressing at all
i think there may be a bug related to that task series.
To milkyWay@Home team: please consider this as a bug report and please investigate the cause of this

Thanks in advance


Next time you get one let it run for a couple of minutes then suspend MW for a slow count of 5 then restart it and see if the processing picks back up. SOMETIMES Boinc forgets what it is doing and stops using the gpu and instead tries to use the cpu to crunch with, the suspend/resume method sometimes brings it back to reality and the gpu kicks back in again.

Profile Marcin
Avatar
Send message
Joined: 18 Jun 09
Posts: 4
Credit: 1,845,212
RAC: 0

Message 58427 - Posted: 28 May 2013, 17:00:47 UTC - in response to Message 58418.



Next time you get one let it run for a couple of minutes then suspend MW for a slow count of 5 then restart it and see if the processing picks back up. SOMETIMES Boinc forgets what it is doing and stops using the gpu and instead tries to use the cpu to crunch with, the suspend/resume method sometimes brings it back to reality and the gpu kicks back in again.

Thanks for the reply.
I will be sure to try the described method


Post to thread

Message boards : Application Code Discussion : N-Body simulation may be faulty over 800 hours computation time estimate


Main page · Your account · Message boards


Copyright © 2017 AstroInformatics Group