KWSN Orbiting Fortress Forum Index KWSN Orbiting Fortress
KWSN Distributed Computing Teams forum
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

Einstein@home optomized app

 
Post new topic   Reply to topic    KWSN Orbiting Fortress Forum Index -> KWSN BOINC'ers
View previous topic :: View next topic  
Author Message
Son Goku
Duke
Duke


Joined: 13 Mar 2006
Posts: 426

PostPosted: Tue Mar 14, 2006 5:44 am    Post subject: Einstein@home optomized app Reply with quote

Noticed this thread earlier which has linked back to the E@H board for an optomized Albert client

http://einstein.phys.uwm.edu/forum_thread.php?id=3532

People are making claims of a 50% reduction in WU completion times, and thus far are claiming their WUs have validated...

There's been a bit of a firestorm of discussion, so the app is towards the beginning of the thread (second post from bottom if viewing new posts first). Supposedly, though there has been no official endorsement, Bruce Allen has supposedly seen the app, and has made no post either suggesting or requesting people don't use it... The full discussion is at the link above...
Back to top
View user's profile Send private message
KWSN - Sir Brian C.......
Stop calling me 'she'
Prince


Joined: 27 Feb 2006
Posts: 2032
Location: Judea, AD33, at a stoning with me mum.

PostPosted: Tue Mar 14, 2006 6:32 am    Post subject: Updated Einstein Optimised executalbe Reply with quote

They've now split the threads so that Akosf ( the guy that tweaked the code) can post his updates

http://einstein.phys.uwm.edu/forum_thread.php?id=3895

still only for x86 only!

Still o official word from the einstein admins

use at your own risk



#ni-1 #ni-1 #ni-1
_________________
Oh, it's blessed are the meek!, Well I'm glad they'll get something as they have a hell of a time!

Back to top
View user's profile Send private message Visit poster's website
Son Goku
Duke
Duke


Joined: 13 Mar 2006
Posts: 426

PostPosted: Tue Mar 14, 2006 6:44 am    Post subject: Reply with quote

Ah yeah, that was due to requests by some that the updates be seperated from the comments. The thread you provided is the update thread with the download, and the thread I provided is the ongoing discussion.

I, think I'm gonna give it a look. Given I'm running some alpha projects already, have been keeping an eye on things, so should be able to catch something from there...

Anyhow, a heads up for whoever might be interested...
Back to top
View user's profile Send private message
Son Goku
Duke
Duke


Joined: 13 Mar 2006
Posts: 426

PostPosted: Tue Mar 14, 2006 9:04 am    Post subject: Reply with quote

27 minutes and 27 seconds with c37 on my Athlon 64 3500+

I'll see how it goes through the validator after more peeps have returned results to get the quorum...

Edit: From what I can see thus far, the validators seem happy, and I'm getting credit...
Back to top
View user's profile Send private message
Son Goku
Duke
Duke


Joined: 13 Mar 2006
Posts: 426

PostPosted: Fri Mar 17, 2006 6:56 am    Post subject: Reply with quote

Looks like this app is good then (and it has validated everything)... Noticed this on the E@H board

http://einstein.phys.uwm.edu/forum_thread.php?id=3532#28234

Quote:
Bruce's promised new, official applications based on akosf's algorithm improvements, but they haven't done so yet. Mainly I suspect the delay is in translating the assembly back into C to allow crossplatform development.


Without looking further, can't say there's a post from Bruce on this, but if he plans on releasing an app which will be official based on this guy's code...
Back to top
View user's profile Send private message
Elwood
Knight
Knight


Joined: 21 Feb 2006
Posts: 84
Location: West Virginia, USA

PostPosted: Mon May 22, 2006 12:55 pm    Post subject: Reply with quote

The reduction in shrub time using the AkosF SSE3 app on a pentium D is nothing short of amazing. A lot of mine have taken 400 seconds or less. Few are taking more than 3,000 secs, down from 20,000 from a comparable shrub pre-optimization. Some of the shorter WU's coming out now are taking me about 400 seconds and claiming a fraction of 1 cobblestone per shrub.

It does create some controversy with the credit claims when using the standard BOINC client, though. One of my larger shrubs claimed 6 and recieved 51. Bad BOINC etiquette or good science efficiency? A question for the ages, I suppose, especially since Trux doesn't have a 5.4 client.

There is also the issue of shrub quota. Einstein only gives 32/day max, so you almost have to have at least one more project on any newer machine running the optimized app, as one can blow through 32 in considerably less than 24 hours with the reduced cruching times.
_________________
Back to top
View user's profile Send private message
Fart in your gen direxion
I am the goatse.cx guy
Prince


Joined: 24 May 2002
Posts: 2022
Location: Regrettably for you, I'm Upwind in Upstate N.Y.

PostPosted: Mon May 22, 2006 1:59 pm    Post subject: Reply with quote

Elwood wrote:
The reduction in shrub time using the AkosF SSE3 app on a pentium D is nothing short of amazing.




All hail the reduction in shrubbing time Cool .

We are all humbled by the magnificence of the shrub.





#ni-2
_________________
Ni ! Ni !

Flatulenty yours,

Sir Fart

The Prince of Noxious Fumigations
The Earl of Eruption
The Baron of Breaking Wind
The Marquis of the Malodorous
The Monarch of Methane
Loony Emeritus
Back to top
View user's profile Send private message Visit poster's website
Son Goku
Duke
Duke


Joined: 13 Mar 2006
Posts: 426

PostPosted: Tue May 23, 2006 6:26 pm    Post subject: Reply with quote

Elwood wrote:
It does create some controversy with the credit claims when using the standard BOINC client, though. One of my larger shrubs claimed 6 and recieved 51. Bad BOINC etiquette or good science efficiency? A question for the ages, I suppose, especially since Trux doesn't have a 5.4 client.


But then there is only so much one can do. Get caught using an optimized CC on Rosseta or QMC, or even Trux's callibrating client, and just see how fast the controversy and claims of cheating will arise. Even if one can turn calibration off (which, well Trux one could but it isn't 5.4.x), there was still talk sometime ago of banning anyone caught using a non-standard CC, and whether some projects should consider this.

One will usually get more then claimed, and yes occassionaly someone will get less due to them not using an optimzed science app also. However, 2 BOINC CCs on the same box and in the same OS install for dif projects, almost defeats the purpose of having a manager, and I for one would rather not enter into the wrong side of this controvery on other projects. Einstein, SETI, and CPDN like projects aren't the only I shrub Very Happy
Back to top
View user's profile Send private message
Elwood
Knight
Knight


Joined: 21 Feb 2006
Posts: 84
Location: West Virginia, USA

PostPosted: Wed May 24, 2006 11:08 pm    Post subject: Reply with quote

Quote:
Without looking further, can't say there's a post from Bruce on this, but if he plans on releasing an app which will be official based on this guy's code...


They are planning on putting the AkosF optimizations into the official application for S5. There is a beta app up now that you can use for S4 crunching if you want.

They are also changing the credit system to something along the lines of Seti-Ehanced or CPDN where credit will be pre-determined based on the size of the WU and will thus ignore benchmarks.
_________________
Back to top
View user's profile Send private message
KWSN - Sir Brian C.......
Stop calling me 'she'
Prince


Joined: 27 Feb 2006
Posts: 2032
Location: Judea, AD33, at a stoning with me mum.

PostPosted: Fri May 26, 2006 3:35 am    Post subject: Reply with quote

Had to upgrade boinc when I got back, seti was all over the place, so now using the unoptimised boinc client, as far as claiming low credit and getting high, you only get the high credit when you're in a quorum with at least two others who are using unoptimised apps. if you're in a quorum with another optimised app user you get the low score.

here's two wu's that show this

high credit grant
http://einstein.phys.uwm.edu/workunit.php?wuid=8822034

low credit grant
http://einstein.phys.uwm.edu/workunit.php?wuid=8806369


here's a link to Akosf's latest clients http://einstein.phys.uwm.edu/forum_thread.php?id=3898



#ni-1 #ni-1
_________________
Oh, it's blessed are the meek!, Well I'm glad they'll get something as they have a hell of a time!

Back to top
View user's profile Send private message Visit poster's website
Son Goku
Duke
Duke


Joined: 13 Mar 2006
Posts: 426

PostPosted: Fri May 26, 2006 3:47 am    Post subject: Reply with quote

Yeah, credit wise I had seen the same, which also makes sense... Yeah I know, we're not supposed to make sense in here #ni-1

Oh, and on the credits, I'm not sure what Einstein will exactly end up doing, though hadn't followed it while I had all that LHC work. Last word I heard, some suggested fpops counting as in SETI, and the impression I got was that some on the Einstein staff were like "well lets not rush into something just yet", perhaps considering some of the talk about whether the stats at SETI now are a bit on the low side, or don't know.

Anyhow, it is their project, so when it all settles, I'm sure Bruce Allen and others will come up with/decide upon something. Perhaps things were more set when I checked last though. Hadn't followed it since about a week ago...
Back to top
View user's profile Send private message
Display posts from previous:   
Post new topic   Reply to topic    KWSN Orbiting Fortress Forum Index -> KWSN BOINC'ers All times are GMT - 5 Hours
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB © 2001, 2005 phpBB Group
Optimized Seti@Home App | BOINC Stats