Are there any sites providing optimized clients? -- PART II

Message boards : Number crunching : Are there any sites providing optimized clients? -- PART II
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 . . . 19 · Next

AuthorMessage
Profile Pappa
Volunteer tester
Avatar

Send message
Joined: 9 Jan 00
Posts: 2562
Credit: 12,301,681
RAC: 0
United States
Message 349563 - Posted: 27 Jun 2006, 1:53:40 UTC - in response to Message 349550.  

Simon

I wish you Luck! If you want a reference they you can give them my lowly al.setiboinc (at) Gmail.com addresss. I will be happy to add references and explain what is happening and how they "could" gain inroads into a "Sun Shop"...

From reading various BOINc Developers Froums, there are inroads into detecting the type of CPU to allow delivoring the correct Application according to the CPU type...

So Progress can be made, IF a Few people Try!

Oi, 649 posts...

And about the donation and related questions...er...no idea, don't really want to comment, nothing good could come of it.

I had a different idea, which is to write to Intel directly (marketing or one the CEO's office) about supporting Seti@Home and giving a license to the project for the express purpose of offering optimized clients for everyone.

Not sure how that will go, I haven't written and sent that letter (yes, snail mail) yet, but I will.

If Intel doesn't want to do that, then I'd try a donation drive towards the same goal - not giving this licence to any one person but to the project.

I really don't believe, even though I spent lots of time doing what needed doing, that I am entitled to any recompensation. Sure, it would be nice, but noone asked me to, right?

I can only hope Intel notices my letter and is interested :o)
After all, there are a lot of us crunchers, and we all need hardware to crunch with...and the project could always use some more server hardware, too, but that's a different thing entirely (and I will not ask them for it in my letter...alluding to it maybe).

Regards,
Simon.


Please consider a Donation to the Seti Project.

ID: 349563 · Report as offensive
Hans Dorn
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 3 Apr 99
Posts: 2262
Credit: 26,448,570
RAC: 0
Germany
Message 349564 - Posted: 27 Jun 2006, 1:53:50 UTC - in response to Message 349551.  



I think the gist of it is not so much the stable versions don't pass muster, but rather they aren't certified for use under the study protocol.
Alinator


I don't know if this argument holds water. IMO their validator should take care of this.

There are other sources for erroneous results:

- faulty hardware
- errors in the official apps

Regards Hans

ID: 349564 · Report as offensive
Profile KWSN - Chicken of Angnor
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 9 Jul 99
Posts: 1199
Credit: 6,615,780
RAC: 0
Austria
Message 349566 - Posted: 27 Jun 2006, 1:55:13 UTC
Last modified: 27 Jun 2006, 1:55:52 UTC

Well,

the whole problem for Akosf and with the Einstein apps is that they're closed source and he had only compiled code to optimize using manual assembly tuning.

The Seti@Home and BOINC clients are both GPL, the Einstein ones aren't. So I'm really not surprised (but sad for Akosf) that they decided to disallow (some of?) his apps.

And yeah, if you distribute a program, you should have a license for it...RTPatch isn't open source or free - but that's never stopped anyone from distributing it, heh...if you have a license, you're allowed to, as well - no idea whether he had one or not. There are free alternatives that are easy to use (even GUI-based).

So anyway, I've always just crunched for Seti, and that won't change anytime soon. No disrespect to other projects, but I'm staying right here :o)

Regards,
Simon.
Donate to SETI@Home via PayPal!

Optimized SETI@Home apps + Information
ID: 349566 · Report as offensive
Profile Pappa
Volunteer tester
Avatar

Send message
Joined: 9 Jan 00
Posts: 2562
Credit: 12,301,681
RAC: 0
United States
Message 349569 - Posted: 27 Jun 2006, 1:57:29 UTC - in response to Message 349559.  

Brian

A portion went into the S5 Application, now he is looking at what is happening and going to work... So If you look and read there are many people working on it... For Seti that work is happening in Seti Beta...

A litlle OT, but Akos had to pull his latest patches because of validation issues.


Perhaps I just didn't sit down and really read what to do with the Einstein stuff, but it just seemed to be more complex than the "Average" user would be able to install. All the patching of .DAT files and whatnot... It shouldn't have been so public...and if patching of files was required then someone should've got an RTPatch license...IMHO.

I didn't bother with an optimized S5. Short run times for me were about 45 minutes with the long ones running at about 7.00 to 7.25 hours..on the stock app.


Please consider a Donation to the Seti Project.

ID: 349569 · Report as offensive
Profile Gecko
Volunteer tester
Avatar

Send message
Joined: 17 Nov 99
Posts: 454
Credit: 6,946,910
RAC: 47
United States
Message 349571 - Posted: 27 Jun 2006, 1:58:06 UTC - in response to Message 349524.  
Last modified: 27 Jun 2006, 2:00:40 UTC

A litlle OT, but Akos had to pull his latest patches because of validation issues.
A discussion about running optimized or not has started in this thread.

Some people are leaving Einstein already.

Regards Hans



A combination of issues. In the testing, some patches indeed produced invalid results and this was "live testing" so those WUs were being uploaded and failing validation.
Biggest issue appears to be that Einstein proj. mgt. only recognizes "official aps" as sanctioned for doing actual science. Unlike S@H, Einstein tightly controls modifications to the applications. Akos's previous S4 opt. aps were released as Official "Betas" after approval of project Mgt. HIs last validating patch, S5T0712 was @ 35% faster than the std application. It would seem there wasn't unanimous approval of his efforts w/ the project and he was requested to tell everyone that "the party was over"....rather abruptly w/o any warning. Needless to say, folks while certainly appreciative and supportive of Akos, are pretty peeved at E@H project Mgt. They're still adjusting to very same things we all went through with migration to Enhanced in May ie. dramatically increased crunching time and credit methodology changes. Many of the comments and discussions on their board right now will look quite familiar to S@H forum participants.
ID: 349571 · Report as offensive
Profile Pappa
Volunteer tester
Avatar

Send message
Joined: 9 Jan 00
Posts: 2562
Credit: 12,301,681
RAC: 0
United States
Message 349575 - Posted: 27 Jun 2006, 2:01:25 UTC - in response to Message 349564.  
Last modified: 27 Jun 2006, 2:02:32 UTC

Hans

Sorry, You forgot one...

Thier Validator with the new Appilication could also be an issue... In Alpha/Beta it is a Bug! After Release it is an "un-documented feature."



I think the gist of it is not so much the stable versions don't pass muster, but rather they aren't certified for use under the study protocol.
Alinator


I don't know if this argument holds water. IMO their validator should take care of this.

There are other sources for erroneous results:

- faulty hardware
- errors in the official apps

Regards Hans


Regards

Pappa

Please consider a Donation to the Seti Project.

ID: 349575 · Report as offensive
Hans Dorn
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 3 Apr 99
Posts: 2262
Credit: 26,448,570
RAC: 0
Germany
Message 349578 - Posted: 27 Jun 2006, 2:06:36 UTC - in response to Message 349555.  

Hans, I believe that too.
Which is why I'm taking the time to try and do that right, or they'll just ignore me.

Regards,
Simon.


While you're at it, could you mention that I want a couple of woodcrests NOW :o)

Regards Hans
ID: 349578 · Report as offensive
Profile KWSN - Chicken of Angnor
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 9 Jul 99
Posts: 1199
Credit: 6,615,780
RAC: 0
Austria
Message 349579 - Posted: 27 Jun 2006, 2:06:43 UTC - in response to Message 349563.  

Simon

I wish you Luck! If you want a reference they you can give them my lowly al.setiboinc (at) Gmail.com addresss. I will be happy to add references and explain what is happening and how they "could" gain inroads into a "Sun Shop"...

From reading various BOINc Developers Froums, there are inroads into detecting the type of CPU to allow delivoring the correct Application according to the CPU type...

So Progress can be made, IF a Few people Try!


Exactly! So that's why I'm trying to spearhead a friendly effort at getting us all to enjoy the potential benefits.

I would be glad if I can include you in my references, thanks for offering. I really don't know what to do first right now, good thing I still have time off from work (anyone looking for a good NT/Linux/Slowaris/BSD sysadmin / web developer / multi-os wizard? *grin*).

Seriously though, I'm really trying hard to make everything I do as transparent as possible to avoid any possible drama.

So far, it's going well :o)

Regards,
Simon.
Donate to SETI@Home via PayPal!

Optimized SETI@Home apps + Information
ID: 349579 · Report as offensive
Profile KWSN - Chicken of Angnor
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 9 Jul 99
Posts: 1199
Credit: 6,615,780
RAC: 0
Austria
Message 349580 - Posted: 27 Jun 2006, 2:08:20 UTC
Last modified: 27 Jun 2006, 2:08:45 UTC

Hans, I believe that too.
Which is why I'm taking the time to try and do that right, or they'll just ignore me.

Regards,
Simon.


While you're at it, could you mention that I want a couple of woodcrests NOW :o)

Regards Hans

Lol.

Yes, yes. A couple of Woodcrests to do the heavy lifting, and a couple of Core 2 Duo X6800s to play with. Er, there goes my mind :o)

Wouldn't we just love that.

Regards,
Simon.
Donate to SETI@Home via PayPal!

Optimized SETI@Home apps + Information
ID: 349580 · Report as offensive
Profile Byron Leigh Hatch @ team Carl Sagan
Volunteer tester
Avatar

Send message
Joined: 5 Jul 99
Posts: 4548
Credit: 35,667,570
RAC: 4
Canada
Message 349581 - Posted: 27 Jun 2006, 2:09:05 UTC - in response to Message 349555.  
Last modified: 27 Jun 2006, 2:22:27 UTC



HI simon

Hi Pappa

Hi Hans




Pappa: Like I said, I believe he paid for them :o) Enough drama.

and Hans, I believe that too.
Which is why I'm taking the time to try and do that right, or they'll just ignore me.

Regards,
Simon.



Some old history from ..... 30 Nov 2005 ... 13:37:48 UTC ...



To be honest i think the windows version will be faster.


But you're talking about your new client, which you won't release,right?



Actually I don't want to buy ipp for $199 and icc for $399.


Crunch3r,

I have a fully licensed copy of MS Visual C++ 2003, but my trial versions of IPP, ICC and the Intel Math Kernel Library just expired as well. I haven't purchased them because I am so new to it, that I couldn't get it to compile without a buttload of error messages and warnings in the compiler output. If you would be willing to release your source code to me, I would be willing to buy a fully licensed copy of ICC & IPP, compile it with your instructions, and send it back to you to release to the public. You and your other source code contributors can take FULL credit for the clients, and I would get to learn a few things in the process. If you would be interested in this arrangement, let me know. If all I would need is the ICC and IPP, I should be able to swing the cost in about 2 weeks. If I would also need the Math Kernel Library, then it will take about 3 or 4 weeks to get the funds.

Regards, Daniel.


That Sounds good to me.

the above was copied from here .............. Posted .. on ... 30 Nov 2005 13:37:48 UTC

http://setiathome.berkeley.edu/forum_thread.php?id=23410#199350


Regards
Byron



ID: 349581 · Report as offensive
Astro
Volunteer tester
Avatar

Send message
Joined: 16 Apr 02
Posts: 8026
Credit: 600,015
RAC: 0
Message 349582 - Posted: 27 Jun 2006, 2:10:39 UTC

The first thread used by TMR when trying to optimize a windows version with the Eval version is called "Compiling faster Windows client with Intel C++ compiler (and fftw?). It's an interesting read. This development sounds much like that one. TMR discusses licenses and such, there are even links to intel compiler forums, and many things that KWSN - Chicken of Angnor might find helpful.

Also, in the follow up thread called 2nd: Compiling faster Windows client with Intel C++ compiler (and fftw?), TMR Identifies the person/s who set him up with the license in this post.

Hope this helps

tony
ID: 349582 · Report as offensive
Profile KWSN - Chicken of Angnor
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 9 Jul 99
Posts: 1199
Credit: 6,615,780
RAC: 0
Austria
Message 349584 - Posted: 27 Jun 2006, 2:14:44 UTC

Thanks a lot for the links!

Lots of useful info. Huge threads, too :o) It'll take me a while to skim over them ...

Regards,
Simon.
Donate to SETI@Home via PayPal!

Optimized SETI@Home apps + Information
ID: 349584 · Report as offensive
Astro
Volunteer tester
Avatar

Send message
Joined: 16 Apr 02
Posts: 8026
Credit: 600,015
RAC: 0
Message 349585 - Posted: 27 Jun 2006, 2:15:11 UTC

ID: 349585 · Report as offensive
Astro
Volunteer tester
Avatar

Send message
Joined: 16 Apr 02
Posts: 8026
Credit: 600,015
RAC: 0
Message 349586 - Posted: 27 Jun 2006, 2:16:32 UTC
Last modified: 27 Jun 2006, 2:16:43 UTC

and shortly after the 3rd thread came, New Linux client compiled with Intel C++ compiler?
ID: 349586 · Report as offensive
Astro
Volunteer tester
Avatar

Send message
Joined: 16 Apr 02
Posts: 8026
Credit: 600,015
RAC: 0
Message 349590 - Posted: 27 Jun 2006, 2:21:48 UTC
Last modified: 27 Jun 2006, 2:22:29 UTC

then, compile a faster linux client using ipp instead of fftw

then, Will you make optimized Windows client if I give the source?,

then, "Legal' optimized client for Windows,

then, Oh heck, just search TMR's post from 383 days ago and sooner, I'll stop the list here, but it continues for a while.

tony


NOTE: these were posted in chronological order of posting
ID: 349590 · Report as offensive
Profile Pappa
Volunteer tester
Avatar

Send message
Joined: 9 Jan 00
Posts: 2562
Credit: 12,301,681
RAC: 0
United States
Message 349592 - Posted: 27 Jun 2006, 2:27:40 UTC - in response to Message 349590.  

Tony

Thank You!

then, compile a faster linux client using ipp instead of fftw

then, Will you make optimized Windows client if I give the source?,

then, "Legal' optimized client for Windows,

then, Oh heck, just search TMR's post from 383 days ago and sooner, I'll stop the list here, but it continues for a while.

tony


NOTE: these were posted in chronological order of posting


Please consider a Donation to the Seti Project.

ID: 349592 · Report as offensive
Profile Pappa
Volunteer tester
Avatar

Send message
Joined: 9 Jan 00
Posts: 2562
Credit: 12,301,681
RAC: 0
United States
Message 349594 - Posted: 27 Jun 2006, 2:28:43 UTC - in response to Message 349581.  

Byron

Thank You!



HI simon

Hi Pappa

Hi Hans




Pappa: Like I said, I believe he paid for them :o) Enough drama.

and Hans, I believe that too.
Which is why I'm taking the time to try and do that right, or they'll just ignore me.

Regards,
Simon.



Some old history from ..... 30 Nov 2005 ... 13:37:48 UTC ...



To be honest i think the windows version will be faster.


But you're talking about your new client, which you won't release,right?



Actually I don't want to buy ipp for $199 and icc for $399.


Crunch3r,

I have a fully licensed copy of MS Visual C++ 2003, but my trial versions of IPP, ICC and the Intel Math Kernel Library just expired as well. I haven't purchased them because I am so new to it, that I couldn't get it to compile without a buttload of error messages and warnings in the compiler output. If you would be willing to release your source code to me, I would be willing to buy a fully licensed copy of ICC & IPP, compile it with your instructions, and send it back to you to release to the public. You and your other source code contributors can take FULL credit for the clients, and I would get to learn a few things in the process. If you would be interested in this arrangement, let me know. If all I would need is the ICC and IPP, I should be able to swing the cost in about 2 weeks. If I would also need the Math Kernel Library, then it will take about 3 or 4 weeks to get the funds.

Regards, Daniel.


That Sounds good to me.

the above was copied from here .............. Posted .. on ... 30 Nov 2005 13:37:48 UTC

http://setiathome.berkeley.edu/forum_thread.php?id=23410#199350


Regards
Byron





Please consider a Donation to the Seti Project.

ID: 349594 · Report as offensive
Astro
Volunteer tester
Avatar

Send message
Joined: 16 Apr 02
Posts: 8026
Credit: 600,015
RAC: 0
Message 349595 - Posted: 27 Jun 2006, 2:29:58 UTC
Last modified: 27 Jun 2006, 2:30:50 UTC

OK one more, a shortcut to the posts of TMR are here. I started from the oldest but when I got to something like post 380 or 390 (383 days ago), I quit working forward
ID: 349595 · Report as offensive
Alinator
Volunteer tester

Send message
Joined: 19 Apr 05
Posts: 4178
Credit: 4,647,982
RAC: 0
United States
Message 349597 - Posted: 27 Jun 2006, 2:33:21 UTC - in response to Message 349564.  
Last modified: 27 Jun 2006, 2:36:44 UTC


I don't know if this argument holds water. IMO their validator should take care of this.

There are other sources for erroneous results:

- faulty hardware
- errors in the official apps

Regards Hans


Very true, but then there is the case where 2 "faulty" test versions get teamed up, or a single FTV gets selected as the canonical. Since they went to an IR of 2, there's a pretty good chance that could happen.

I think they just want to limit the number of variables present when they post process the MSD at the end of the S5 run.

Alinator
ID: 349597 · Report as offensive
Hans Dorn
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 3 Apr 99
Posts: 2262
Credit: 26,448,570
RAC: 0
Germany
Message 349605 - Posted: 27 Jun 2006, 2:39:07 UTC - in response to Message 349597.  


Very true, but then there is the case where 2 "faulty" test versions get teamed up, or a single FTV gets selected as the canonical. Since they went to an IR of 2, there's a pretty good chance that could happen.

I think they just want to limit the number of variables present when they post process the MSD at the end of the S5 run.

Alinator


Yep. With an IR of 2 the chances are pretty high that bad results get validated.
It's a pity that there's no offline validator for Einstein.

Regards Hans
ID: 349605 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 . . . 19 · Next

Message boards : Number crunching : Are there any sites providing optimized clients? -- PART II


 
©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.