// DBOINCP-300: added node comment count condition in order to get Preview working ?>
dougdoug
Joined: 22 Oct 07
Posts: 1
Credit: 76522
RAC: 0
23 Dec 2007 12:24:41 UTC
Topic 193393
(moderation:
)
I have been running seti for several years and have a nice certificate from the Planetary Society.
I wish there was the same for Einstein. {This is a wish list site}
I just noticed this evening that there is now certificates (account, team, & cross-project) available on the account page. I´m assuming this came to being with the most recent project BOINC web interface upgrade.
My question is, how accurate are the numbers reported? (cobblestones of computation equaling X amount of floating-point operations) Taking into account how ¨fluid¨ (for want of a better word) credit is.
My question is, how accurate are the numbers reported? (cobblestones of computation equaling X amount of floating-point operations) Taking into account how ¨fluid¨ (for want of a better word) credit is.
I really would not take the "floating point operations" at face value, it's more a rough guesstimation. When you think about it, it's even hard to define what a floating point operation exactly is...
OK, if you add, or multiply two floating point numbers, I guess that should count as one operation. Changing the sign of a fp number? If you subtract by chnaging the sign and then adding, does this make one or two operations???
Does an operation that transports a floating point number from/into memory count as well? What about vector operations that operate on several floating point numbers in parallel.... and so on.
Bikeman, thanks for putting that in perspective. So I guess, if I printed a copy and framed it on the wall and someone asks me about it I can say. ¨It means my computers have been stress tested quite extensively." :D
SETIficate
)
I just noticed this evening that there is now certificates (account, team, & cross-project) available on the account page. I´m assuming this came to being with the most recent project BOINC web interface upgrade.
My question is, how accurate are the numbers reported? (cobblestones of computation equaling X amount of floating-point operations) Taking into account how ¨fluid¨ (for want of a better word) credit is.
RE: My question is, how
)
I really would not take the "floating point operations" at face value, it's more a rough guesstimation. When you think about it, it's even hard to define what a floating point operation exactly is...
OK, if you add, or multiply two floating point numbers, I guess that should count as one operation. Changing the sign of a fp number? If you subtract by chnaging the sign and then adding, does this make one or two operations???
Does an operation that transports a floating point number from/into memory count as well? What about vector operations that operate on several floating point numbers in parallel.... and so on.
CU
Bikeman
Bikeman, thanks for putting
)
Bikeman, thanks for putting that in perspective. So I guess, if I printed a copy and framed it on the wall and someone asks me about it I can say. ¨It means my computers have been stress tested quite extensively." :D