New Einstein application!

Message boards : Number crunching : New Einstein application!
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile MattDavis
Volunteer tester
Avatar

Send message
Joined: 11 Nov 99
Posts: 919
Credit: 934,161
RAC: 0
United States
Message 397883 - Posted: 15 Aug 2006, 4:39:04 UTC

Einstein@home just released a new, project-supported optimized app that cuts crunch times by 1/3! This isn't 3rd party - it's downloaded automatically though BOINC!

Just a heads up since people here like to talk about optimized apps.
-----
ID: 397883 · Report as offensive
Profile Ocean Archer
Volunteer tester
Avatar

Send message
Joined: 11 Jun 04
Posts: 148
Credit: 504,219
RAC: 0
United States
Message 398048 - Posted: 15 Aug 2006, 12:34:20 UTC

Is this an app that is downloaded automatically, or do you have to make a specific request to obtain it?
ID: 398048 · Report as offensive
CJOrtega

Send message
Joined: 15 May 99
Posts: 186
Credit: 1,126,273
RAC: 0
United States
Message 398055 - Posted: 15 Aug 2006, 12:46:15 UTC - in response to Message 398048.  

Is this an app that is downloaded automatically, or do you have to make a specific request to obtain it?



It's automatic. My systems have been crunching with them overnight.
ID: 398055 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 398056 - Posted: 15 Aug 2006, 12:47:11 UTC - in response to Message 398048.  

It's essentially the same as the Beta application. If you are running the Beta application, exit BOINC, navigate to your BOINC\\projects\\einstein.phys.uwm.edu\\ folder and delete the app_info.xml file. Restart BOINC. Einstein will continue where it left off.
ID: 398056 · Report as offensive
Pepperammi

Send message
Joined: 3 Apr 99
Posts: 200
Credit: 737,775
RAC: 0
United Kingdom
Message 398106 - Posted: 15 Aug 2006, 14:48:29 UTC
Last modified: 15 Aug 2006, 14:51:58 UTC

Its been a beta for a while I believe but for a little while its been possible to get via the forums and use in the open (officially from Bernd). Sometime yesturday or the day before it was made 'public' (best word i can use) and is now part of the automatic update service. You don't have to do anything (unless your using an appinfo.xml to use the beta version) it'll download itself with new WUs. Have to say its by far faster :)
"It features a codebase similar to the Linux 4.17 Beta Test App, with hand-coded optimized assembler calculation from Akos (and me)."

brilliant work from Bernd and Akos.
ID: 398106 · Report as offensive
Profile RandyC
Avatar

Send message
Joined: 20 Oct 99
Posts: 714
Credit: 1,704,345
RAC: 0
United States
Message 398154 - Posted: 15 Aug 2006, 16:18:14 UTC - in response to Message 397883.  

Einstein@home just released a new, project-supported optimized app that cuts crunch times by 1/3! This isn't 3rd party - it's downloaded automatically though BOINC!

Just a heads up since people here like to talk about optimized apps.


Along with reducing crunch times, they are adjusting granted credit too.

http://einstein.phys.uwm.edu/forum_thread.php?id=4640&nowrap=true#45029

http://einstein.phys.uwm.edu/forum_thread.php?id=4674
ID: 398154 · Report as offensive
Profile michael37
Avatar

Send message
Joined: 23 Jul 99
Posts: 311
Credit: 6,955,447
RAC: 0
United States
Message 399114 - Posted: 16 Aug 2006, 3:26:00 UTC

Is Einstein@home open source or not? Is their application optimized for SSE/SSE2/SSE3/SSE4 for the newer processors such as Seti Enhanced?

If not, what's the point of running a scientific application which cannot take full advantage of your CPUs due to developer shortsightedness?


ID: 399114 · Report as offensive
Tony DeBari

Send message
Joined: 31 Mar 01
Posts: 29
Credit: 14,006,420
RAC: 0
United States
Message 399175 - Posted: 16 Aug 2006, 6:20:41 UTC - in response to Message 399114.  

Is Einstein@home open source or not? Is their application optimized for SSE/SSE2/SSE3/SSE4 for the newer processors such as Seti Enhanced?

The Einstein@Home science app is not open source. IIRC, it can detect an SSE-capable processor and will utilize SSE-specific code if one is found. It cannot as yet take advantage of the more advanced SSEx instruction sets.

If not, what's the point of running a scientific application which cannot take full advantage of your CPUs due to developer shortsightedness?

Developer shortsightedness?? Excuse me, but the official Seti@Home Enhanced science app can't even detect and utilize SSE, let alone anything more advanced. Such is the compromise between producing a fast application and one that will run on as many CPU types as possible out of the box. Simon is doing great work optimizing the current code base for every flavor of CPU he can get his hands on, but his apps and 'cook books' are not official (though they are endorsed by the project).

The rules are different over at Einstein - optimized apps have to be released by the project devs or the results are deemed unacceptable by the project scientists. The latest release of the Einstein S5 app is upwards of 33% faster that the original version, and since it's an officially released app, you don't need an app_info.xml file to use it - every machine will download it automatically. Work on further optimization continues, but I doubt you'll see processor-specific apps (a la Simon's) at Einstein any time soon.

Regards,

-- Tony
ID: 399175 · Report as offensive
Odysseus
Volunteer tester
Avatar

Send message
Joined: 26 Jul 99
Posts: 1808
Credit: 6,701,347
RAC: 6
Canada
Message 399184 - Posted: 16 Aug 2006, 7:45:27 UTC - in response to Message 399175.  

Work on further optimization continues, but I doubt you'll see processor-specific apps (a la Simon's) at Einstein any time soon.

There are already processor-specific apps being tested, for Linux on DEC Alpha and for Mac OS 10.4 on G5.
ID: 399184 · Report as offensive
HTH
Volunteer tester

Send message
Joined: 8 Jul 00
Posts: 691
Credit: 909,237
RAC: 0
Finland
Message 399195 - Posted: 16 Aug 2006, 9:03:55 UTC - in response to Message 399175.  

Excuse me, but the official Seti@Home Enhanced science app can't even detect and utilize SSE, let alone anything more advanced.


What? I cannot believe this!

You are probably speaking about some old version of Enhanced client...

Manned mission to Mars in 2019 Petition <-- Sign this, please.
ID: 399195 · Report as offensive
W-K 666 Project Donor
Volunteer tester

Send message
Joined: 18 May 99
Posts: 19733
Credit: 40,757,560
RAC: 67
United Kingdom
Message 399380 - Posted: 16 Aug 2006, 15:24:43 UTC - in response to Message 399114.  
Last modified: 16 Aug 2006, 15:25:25 UTC

Is Einstein@home open source or not? Is their application optimized for SSE/SSE2/SSE3/SSE4 for the newer processors such as Seti Enhanced?

If not, what's the point of running a scientific application which cannot take full advantage of your CPUs due to developer shortsightedness?

There is an explaination about the Einstein app in post MS Windows Beta Test App 4.24 available by Bernd Machenschalk the developer.
In reply to questions by Crunch3r.

Andy
ID: 399380 · Report as offensive
Josef W. Segur
Volunteer developer
Volunteer tester

Send message
Joined: 30 Oct 99
Posts: 4504
Credit: 1,414,761
RAC: 0
United States
Message 399397 - Posted: 16 Aug 2006, 16:34:34 UTC - in response to Message 399195.  
Last modified: 16 Aug 2006, 16:35:49 UTC

Excuse me, but the official Seti@Home Enhanced science app can't even detect and utilize SSE, let alone anything more advanced.


What? I cannot believe this!

You are probably speaking about some old version of Enhanced client...

The FFTW3 code does adapt to various SIMD types, so SSE is used for doing FFTs if available. That's the only adaptable part in current stock releases.

In the current Beta source, Eric has added routines which can test various functions for:

Baseline smoothing
Converting FFT output to a PowerSpectrum
Transposing the PowerSpectrum to a cache-friendly form
Chirping

Preprocessor definitions determine which variations are compiled into the application, run-time testing determines which variations can be used by the CPU and chooses the fastest one.

However, other than Altivec functions Alex Kan supplied there are no SIMD alternatives in the source. If someone would supply those, I think Eric would add them ASAP. They'd need to be checked for correct results on systems which enable each variation.
                                                                Joe
ID: 399397 · Report as offensive
HTH
Volunteer tester

Send message
Joined: 8 Jul 00
Posts: 691
Credit: 909,237
RAC: 0
Finland
Message 401332 - Posted: 19 Aug 2006, 12:04:14 UTC - in response to Message 399397.  

However, other than Altivec functions Alex Kan supplied there are no SIMD alternatives in the source. If someone would supply those, I think Eric would add them ASAP. They'd need to be checked for correct results on systems which enable each variation.


How much will they (theoretically) speed up the crunching? I have Pentium D 920, which has SSE, SSE2 and SSE3.

When will they use SIMD in SETI@home?

Manned mission to Mars in 2019 Petition <-- Sign this, please.
ID: 401332 · Report as offensive
Josef W. Segur
Volunteer developer
Volunteer tester

Send message
Joined: 30 Oct 99
Posts: 4504
Credit: 1,414,761
RAC: 0
United States
Message 401521 - Posted: 19 Aug 2006, 17:15:44 UTC - in response to Message 401332.  

However, other than Altivec functions Alex Kan supplied there are no SIMD alternatives in the source. If someone would supply those, I think Eric would add them ASAP. They'd need to be checked for correct results on systems which enable each variation.


How much will they (theoretically) speed up the crunching? I have Pentium D 920, which has SSE, SSE2 and SSE3.


On Simon Zadra's site, this thread includes some profiling data by Ben Herndon. The chirping and conversion to PowerSpectrum together account for slightly more than 5 percent of crunch time. Optimized versions might save most of that time, and since the conversion to PowerSpectrum is used in baseline smoothing, that would also be faster. Ben is working from Simon's 1.3 sources AFAIK, so transposition of the PowerSpectrum isn't included.

When will they use SIMD in SETI@home?

When someone submits the needed routines. It's an open source project, Eric Korpela is already paddling his canoe as fast as he is able, so someone else will have to help out in order for those routines to be included in the current Beta cycle.
                                                          Joe
ID: 401521 · Report as offensive

Message boards : Number crunching : New Einstein application!


 
©2025 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.