Message boards :
Number crunching :
Validation Process
Message board moderation
Author | Message |
---|---|
ChrisD Send message Joined: 25 Sep 99 Posts: 158 Credit: 2,496,342 RAC: 0 |
How accurate must a returned result be to be accepted? Please do not laugh, I am serious. From posts in the programmer's threads, I know that each client produces slightly different results depending on which processor does the processing. MMX-SSE stuff, Linux - Windows - Mac You know. To check things out, I am right now trying to make 2 of my machines crunch the same WU. One Athlon XP and an Intel PII MMX, just to see if they return exactly the same figures, or they differ in the 7th or 8th digit. However, if the 'Coding people' reads this, they might save me the hazzle by telling me right away. :) ChrisD |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
> How accurate must a returned result be to be accepted? > > Please do not laugh, I am serious. > > From posts in the programmer's threads, I know that each client produces > slightly different results depending on which processor does the processing. > MMX-SSE stuff, Linux - Windows - Mac You know. > > To check things out, I am right now trying to make 2 of my machines crunch the > same WU. One Athlon XP and an Intel PII MMX, just to see if they return > exactly the same figures, or they differ in the 7th or 8th digit. For SETI@Home, you will get minor fluctuations because of the FPUs not delivering the precicely identical values. What the validator is going to do is check to see if you found the same spikes, Gaussians, etc. in the work unit that the other participants did when they processed that same WU. If you find the same number of spikes but yours is less than, or greater than, theirs by some %, depending on how large that error is will either say it was valid, or reject it. Like wise with spike and Gaussian detection if you "find" some they don't it is more than likely more than a rounding problem. The vaildation is project dependent and can be very strict (LHC@Home has this problem). <p> For BOINC Documentaion: Click Me! |
Scott Brown Send message Joined: 5 Sep 00 Posts: 110 Credit: 59,739 RAC: 0 |
> The vaildation is project dependent and can be very strict (LHC@Home has this > problem). I think that with the new release of the LHC client, this issue has been fixed. |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
|
©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.