Welcome to MilkyWay@home

new work (4/02)

Message boards : Number crunching : new work (4/02)
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 2953 - Posted: 2 Apr 2008, 17:09:53 UTC

I know we're out of work. I'm running a purge right now and there should be more work available as soon as it finishes :)
ID: 2953 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 2955 - Posted: 2 Apr 2008, 19:20:29 UTC - in response to Message 2953.  

I know we're out of work. I'm running a purge right now and there should be more work available as soon as it finishes :)


i've started up a new (really big) batch of searches. these should keep you happily crunching for a long time (i hope) :D
ID: 2955 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile nickth
Avatar

Send message
Joined: 5 Oct 07
Posts: 33
Credit: 3,189,992
RAC: 0
Message 2957 - Posted: 2 Apr 2008, 19:32:58 UTC - in response to Message 2955.  

I know we're out of work. I'm running a purge right now and there should be more work available as soon as it finishes :)


i've started up a new (really big) batch of searches. these should keep you happily crunching for a long time (i hope) :D



Thanks Travis
Whats this new file stars_convolved_82_simulated.txt its big anyway at 6mb

Smoke me a kipper ill be back for breakfast


ID: 2957 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 2959 - Posted: 2 Apr 2008, 19:47:35 UTC - in response to Message 2957.  

I know we're out of work. I'm running a purge right now and there should be more work available as soon as it finishes :)


i've started up a new (really big) batch of searches. these should keep you happily crunching for a long time (i hope) :D



Thanks Travis
Whats this new file stars_convolved_82_simulated.txt its big anyway at 6mb


it's a new set of star points to crunch on. Nate has been running into some problems with using conjugate gradient descent on different data sets (its converging to a local minima), and interestingly enough we were getting the same problem with the genetic search. we're trying out a new set of data, and tweaked our genetic search a little bit (we think some of the parameters in the model aren't orthogonal -- which is screwing things up). so these new searches on this new data should help us figure out the problem.
ID: 2959 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile nickth
Avatar

Send message
Joined: 5 Oct 07
Posts: 33
Credit: 3,189,992
RAC: 0
Message 2960 - Posted: 2 Apr 2008, 19:56:01 UTC - in response to Message 2959.  

I know we're out of work. I'm running a purge right now and there should be more work available as soon as it finishes :)


i've started up a new (really big) batch of searches. these should keep you happily crunching for a long time (i hope) :D



Thanks Travis
Whats this new file stars_convolved_82_simulated.txt its big anyway at 6mb


it's a new set of star points to crunch on. Nate has been running into some problems with using conjugate gradient descent on different data sets (its converging to a local minima), and interestingly enough we were getting the same problem with the genetic search. we're trying out a new set of data, and tweaked our genetic search a little bit (we think some of the parameters in the model aren't orthogonal -- which is screwing things up). so these new searches on this new data should help us figure out the problem.


Well lets hope it works.

Also what happening about sortting the server out so that you can run automatic purges instead of having to do it mannualy yhus giving you more time to sort out the searches which will inturn give us more work without the breaks for purging

Smoke me a kipper ill be back for breakfast


ID: 2960 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Philadelphia
Avatar

Send message
Joined: 9 Nov 07
Posts: 131
Credit: 180,454
RAC: 0
Message 2963 - Posted: 2 Apr 2008, 20:24:11 UTC
Last modified: 2 Apr 2008, 20:25:41 UTC

Thanks for the new WU's :)

PS- Is it me or are the WU's getting longer?
CLICK TO HELP BUILD
ID: 2963 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Cori
Avatar

Send message
Joined: 27 Aug 07
Posts: 647
Credit: 27,592,547
RAC: 0
Message 2964 - Posted: 2 Apr 2008, 20:42:09 UTC - in response to Message 2963.  

... PS- Is it me or are the WU's getting longer?

Hm, maybe your computers are just tired and need a rest? *grin*
To me the WUs seem to be of the same length as before. ;-)
Lovely greetings, Cori
ID: 2964 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile banditwolf
Avatar

Send message
Joined: 12 Nov 07
Posts: 2425
Credit: 524,164
RAC: 0
Message 2965 - Posted: 2 Apr 2008, 20:54:07 UTC

Seem the same too.
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.
ID: 2965 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
[AF>EDLS>BIOMED] Heyoka

Send message
Joined: 29 Aug 07
Posts: 1
Credit: 68,743
RAC: 0
Message 2966 - Posted: 2 Apr 2008, 21:57:28 UTC - in response to Message 2953.  

I know we're out of work. I'm running a purge right now and there should be more work available as soon as it finishes :)


And if you reduce the deadline, one or two day.
It could be better ?
ID: 2966 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile DoctorNow
Avatar

Send message
Joined: 28 Aug 07
Posts: 146
Credit: 10,544,989
RAC: 9,645
Message 2969 - Posted: 3 Apr 2008, 4:56:31 UTC - in response to Message 2963.  

PS- Is it me or are the WU's getting longer?

I have the same feeling, but only a wee bit, some seconds maybe.
Member of BOINC@Heidelberg and ATA!

My BOINCstats
ID: 2969 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 2971 - Posted: 3 Apr 2008, 5:36:22 UTC - in response to Message 2969.  

PS- Is it me or are the WU's getting longer?

I have the same feeling, but only a wee bit, some seconds maybe.


the WUs might be a little longer, because the star file is bigger. more stars to calculate the likelihood would increase the time of calculation.

basically, the WU does two parts, the first calculates an integral over the volume of space we're dealing with. this is pretty much fixed -- we can set how fine grained it is, but once thats fixed it should take the same amount of time.

the second part calculates the likelihood of the calculated model being a good match to the data. this is usually pretty quick as it's done in linear time over the stars in the stars file, and not nearly as computationally intensive as calculating the integral. so when we increase the star file you should see the WUs take slightly longer.

this is also the reason that the progress bar is a bit wonky, when it changes speeds that means it's stopped calculating the integral and started to calculate the likelihood.
ID: 2971 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile DoctorNow
Avatar

Send message
Joined: 28 Aug 07
Posts: 146
Credit: 10,544,989
RAC: 9,645
Message 2972 - Posted: 3 Apr 2008, 7:24:55 UTC

Thanks for the explanation, Travis. :-)
Member of BOINC@Heidelberg and ATA!

My BOINCstats
ID: 2972 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Philadelphia
Avatar

Send message
Joined: 9 Nov 07
Posts: 131
Credit: 180,454
RAC: 0
Message 2987 - Posted: 3 Apr 2008, 22:33:59 UTC - in response to Message 2971.  

PS- Is it me or are the WU's getting longer?

I have the same feeling, but only a wee bit, some seconds maybe.


the WUs might be a little longer, because the star file is bigger. more stars to calculate the likelihood would increase the time of calculation.

basically, the WU does two parts, the first calculates an integral over the volume of space we're dealing with. this is pretty much fixed -- we can set how fine grained it is, but once thats fixed it should take the same amount of time.

the second part calculates the likelihood of the calculated model being a good match to the data. this is usually pretty quick as it's done in linear time over the stars in the stars file, and not nearly as computationally intensive as calculating the integral. so when we increase the star file you should see the WUs take slightly longer.

this is also the reason that the progress bar is a bit wonky, when it changes speeds that means it's stopped calculating the integral and started to calculate the likelihood.


Thanks Travis.

CLICK TO HELP BUILD
ID: 2987 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Paratima

Send message
Joined: 15 Nov 07
Posts: 31
Credit: 56,404,447
RAC: 0
Message 2988 - Posted: 3 Apr 2008, 23:30:09 UTC
Last modified: 3 Apr 2008, 23:30:58 UTC

WU's on my main crunchers are running exactly the same as before, 5:07, on a 3GHz AMD, according to the BoincView monitor. I'd say your PC needs a vacation.
ID: 2988 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile niterobin

Send message
Joined: 11 Mar 08
Posts: 28
Credit: 818,194
RAC: 0
Message 3002 - Posted: 4 Apr 2008, 5:04:42 UTC

On my 1200 MHz Duron workunits are 15 seconds faster at 13m 44s. On my 2.4 GHz P4, they're 5 seconds faster at 12m 24s. So, thank you for the speed increase. :-)
ID: 3002 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Emanuel

Send message
Joined: 18 Nov 07
Posts: 280
Credit: 2,442,757
RAC: 0
Message 3010 - Posted: 4 Apr 2008, 14:01:08 UTC - in response to Message 3002.  

There could certainly be some truth to this. Cache misses and consequent memory performance are hard to predict, not to mention branch predictors failing et cetera. However, I've seen no difference on this PC aside from some differences between different WUs (i.e. different gs_5xx)
ID: 3010 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Number crunching : new work (4/02)

©2024 Astroinformatics Group