Credit: How is it calculated?

Message boards : Number crunching : Credit: How is it calculated?
Message board moderation

To post messages, you must log in.

AuthorMessage
Jonathan

Send message
Joined: 27 Jun 15
Posts: 4
Credit: 680,121
RAC: 0
United States
Message 1706371 - Posted: 29 Jul 2015, 18:48:09 UTC

I'm curious. What is credit based off of, how is it calculated, and how is recent average credit calculated?
ID: 1706371 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1706376 - Posted: 29 Jul 2015, 18:53:45 UTC - in response to Message 1706371.  

Look here:

https://boinc.berkeley.edu/wiki/computation_credit

and here:

https://boinc.berkeley.edu/trac/wiki/CreditNew
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1706376 · Report as offensive
OzzFan Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Apr 02
Posts: 15691
Credit: 84,761,841
RAC: 28
United States
Message 1706467 - Posted: 29 Jul 2015, 22:14:47 UTC - in response to Message 1706371.  

When you figure it out, let the rest of us know! :-P
ID: 1706467 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1706474 - Posted: 29 Jul 2015, 22:32:21 UTC - in response to Message 1706467.  
Last modified: 29 Jul 2015, 22:38:13 UTC

When you figure it out, let the rest of us know! :-P


I can do that, because it's surprisingly simple when you boil it down.

[tongue firmly in cheek of course]
You start with a natural stochastic process and plug it into an averaging filter that has quantisation noise, add some temporal mismatch, which amounts to topological mixing, creating dense periodic orbits around a rate proportional to the average turnaround time [actually time to validation events]. Then you apply a downscaling (normalisation) related to the proportion of AVX, SSE1-4, and non SIMD reporting clients, and average with your wingman, and apply the cobblestone scale against the efficiency estimate compared to initial estimates of floating point operations (which tend to be relatively reasonable [on this project, for MB, not sure about AP]).

See ? wasn't that hard.
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1706474 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1706475 - Posted: 29 Jul 2015, 22:37:19 UTC - in response to Message 1706474.  

I'm here for the toaster :P
ID: 1706475 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1706484 - Posted: 29 Jul 2015, 22:54:57 UTC - in response to Message 1706474.  

[tongue firmly in cheek of course]
You start with a natural stochastic process and plug it into an averaging filter that has quantisation noise, add some temporal mismatch, which amounts to topological mixing, creating dense periodic orbits around a rate proportional to the average turnaround time [actually time to validation events]. Then you apply a downscaling (normalisation) related to the proportion of AVX, SSE1-4, and non SIMD reporting clients, and average with your wingman, and apply the cobblestone scale against the efficiency estimate compared to initial estimates of floating point operations (which tend to be relatively reasonable [on this project, for MB, not sure about AP]).

Can you please give the technical explanation next? That normal people's jargon is hard to follow for us tech-heads. :(
ID: 1706484 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1706485 - Posted: 29 Jul 2015, 22:57:35 UTC - in response to Message 1706484.  
Last modified: 29 Jul 2015, 23:01:06 UTC

[tongue firmly in cheek of course]
You start with a natural stochastic process and plug it into an averaging filter that has quantisation noise, add some temporal mismatch, which amounts to topological mixing, creating dense periodic orbits around a rate proportional to the average turnaround time [actually time to validation events]. Then you apply a downscaling (normalisation) related to the proportion of AVX, SSE1-4, and non SIMD reporting clients, and average with your wingman, and apply the cobblestone scale against the efficiency estimate compared to initial estimates of floating point operations (which tend to be relatively reasonable [on this project, for MB, not sure about AP]).

Can you please give the technical explanation next? That normal people's jargon is hard to follow for us tech-heads. :(


Haha, I could, but that would take all the fun out, and I have a dentist appointment later today :P *exits stage right*

[Edit:] I do have a dodgy diagram somewhere... hmmm, wonder where it went, oh well later.
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1706485 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1706486 - Posted: 29 Jul 2015, 23:00:11 UTC

Credit: How is it calculated?


Randomly, and poorly.
ID: 1706486 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1706489 - Posted: 29 Jul 2015, 23:06:01 UTC - in response to Message 1706485.  

[Edit:] I do have a dodgy diagram somewhere... hmmm, wonder where it went, oh well later.

Yes! A flow chart, please. :)
ID: 1706489 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1706492 - Posted: 29 Jul 2015, 23:13:28 UTC - in response to Message 1706489.  
Last modified: 29 Jul 2015, 23:23:28 UTC

[Edit:] I do have a dodgy diagram somewhere... hmmm, wonder where it went, oh well later.

Yes! A flow chart, please. :)


Been quite a while since I looked at it, but if it was in my typical style it'll be like a flowchart in a bus accident with a circuit diagram.

Incidentally no bugs were found in a pretty lengthy investigation. A number of design limits do create some weird interactions, but instabilities/limits aside the concept and logic is there. Just needs a bit of refinement.

[Edit:] The two feedback averaging triangles probably aren't hooked up right, but illustrate the concept anyway

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1706492 · Report as offensive
Profile betreger Project Donor
Avatar

Send message
Joined: 29 Jun 99
Posts: 11360
Credit: 29,581,041
RAC: 66
United States
Message 1706554 - Posted: 30 Jul 2015, 1:45:08 UTC - in response to Message 1706486.  

Credit: How is it calculated?


Randomly, and poorly.

Yes, but it does add an added sense of adventure.
ID: 1706554 · Report as offensive
Profile Gary Charpentier Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 25 Dec 00
Posts: 30608
Credit: 53,134,872
RAC: 32
United States
Message 1706581 - Posted: 30 Jul 2015, 5:30:41 UTC - in response to Message 1706371.  
Last modified: 30 Jul 2015, 5:35:57 UTC

I'm curious. What is credit based off of, how is it calculated, and how is recent average credit calculated?

/dev/ranrom * credit_screw(# pulses)
ID: 1706581 · Report as offensive
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1706592 - Posted: 30 Jul 2015, 6:29:22 UTC - in response to Message 1706581.  
Last modified: 30 Jul 2015, 7:24:08 UTC

Well that's the effect from the users' perspectives, but there are others. As the dodgy diagram very loosely shows, it does make a more or less complete closed loop control system. That there is opportunity for extra nice things left out, that the filters overlap in time (fight one another), and the apparent randomness isn't damped using a controller, you could argue would just be tuning from an engineering standpoint.

The main missing bits are to use damped controllers instead of sample averages for the feedback, that the separate averages would need to be made to operate over different timespans (e.g. host over 1-10 tasks, app version over a day or so, and that then [missing] app and project averages could operate slower than that.

For hosts and appversions, extended kalman filters might have been better choices than sample averages, then for apps(multiple app versions) and project level *multiple apps) perhaps particle filters.

There is some scaling error due to using non-SIMD bionic whetstone, that can result in an incorrect figure of > 100% efficiency on AVX equipped hosts+apps, but coarse correction there would likely be much better& simpler than current.

What can fall out of the above corrections, is a whole heap of statistical logic that can be applied back to the initial estimates, and all sorts of sanity/error/cheat prevention, not to mention that if working could provide a path to cross project parity that isn't a pipe-dream. That starts to move out of not funded two man development into large scale engineering though (being realistic), and amounts to AI (expert systems)
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1706592 · Report as offensive

Message boards : Number crunching : Credit: How is it calculated?


 
©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.