Slartibartfast 2003-10-06 03:29:21 | Hi there, I'd be new to this project, one of my PC's has generated a bunch of results already. There's something I haven't figured out yet. On what merit do results get rechecked? To me it looks like a result taking more than about 250Mpts to run gets quarantined and then run 5 times. So far all of them have passed quarantine, but another I wonder about is how come the 5 runs can have different intermediate results and still pass? excerpt from the 1st to last job, another one like that: sfinall=999;sfinalf=999;#runs=5; 9.153511 (1686.2 Mpts) [v4.32b] {132DDC59} the 5 runs in this job had differing values for the amount of timesteps it took to complete. I wonder what that's about, since if it's rechecking the result's validity, shouldn't it be run again with the same start parameters and hence come up with the same values (assuming your CPU or memory isn't shot). Anyway, looking rather well. Submitted the 30 odd jobs, of which 5 or so are in the best100.txt of 9am today already, this one looks to be making it as well once submitted. And the 3rd best Muon % in Solenoids to 15cm, that's a first as well in DC projects that I ran. (of course other DC projects didn't compute designs for particle accelerators, so it makes sense I never had Muon %'s making it to the 3rd spot ) Slarti. |
[DPC]Stephan202 2003-10-06 10:34:26 | Hi, Quarantined results are results that yield higher than any result in results.dat. They are always rechecked 4 times (making a total of five checks) and then the average yield and total Mpts are stored in results.dat and results.txt The reason every run has a different outcome, is that a part of the simulation depends on randomness, like with real-life particles. Oh. And you are a member of the Dutch Power Cows, right? --- Dutch Power Cow. MOOH! |
Slartibartfast 2003-10-06 11:29:58 | ah, cheers for the heads up. I was wondering why it kept doing that (especially before I started to use the best100.txt). Actually, I'm not with any team at all on this project, flying solo. |