Message boards :
Number crunching :
Workunit validated while tasks have different results
Message board moderation
Author | Message |
---|---|
BetelgeuseFive Send message Joined: 6 Jul 99 Posts: 158 Credit: 17,117,787 RAC: 19 |
Hello folks, I just checked some of my tasks and noticed the following workunit: http://setiathome.berkeley.edu/workunit.php?wuid=1510088986 My wingman reported 30 single pulses and 30 repetitive pulses while my task showed 0/0. However both tasks were marked as valid and received credit. How can this happen ? Tom |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
The stderr_txt is not the result file. Before the other result found the pulses and exited its results must have been similar to your results to be declared a match. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
BetelgeuseFive Send message Joined: 6 Jul 99 Posts: 158 Credit: 17,117,787 RAC: 19 |
The stderr_txt is not the result file. Thanks for the info, but it does not explain how a result that is obviously not correct (my wingman has hundreds of invalids and only two valids: this task and a task that was 100 percent blanked) could be used to validate something. The reason I looked at the result in more detail was that I would like to know what kind of GPU could do an Astropulse 25 times faster than my GTX-750. Is there no mechanism to check if tasks that found no pulses are correct (e.g. reporting the closest thing to a pulse that was found ?). Tom |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
The stderr_txt is not the result file. I think the pulses the stock app found are false pulses that the lunatics app sees are false and doesn't report them. Perhaps it was the MB app where that was mentioned. I can't recall at the moment. There are several guys that have an intimate knowledge of how the apps and matching system work. They can give you a detailed answer as to how the results match. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
The stderr_txt is not the result file. I regret not seeing this before the details were purged, and if anyone happened to save some more info please post it. Even just a link to the host with a lot of errors would be interesting. The AP validation has some quirks, but if the uploaded result files had the indicated number of signals there's no way they should both have gotten credit. The only thing I can think of which might explain it is that the 30/30 signals might have had Inf or NaN values which confused the comparisons. Proper processing of AP tasks never gets anywhere near numerical limits, but anything can happen. Because the stock OpenCL GPU apps for AP were provided by Lunatics, the only differences from optimized are what code changes were made. IIRC the stock builds were from SVN revision 1831 while Lunatics v0.41 installer has builds from revision 1843. In any case, there are many users of each being very successful though the apps have not achieved immunity from Murphy's Law. Joe |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.