Welcome to MilkyWay@home

Reverting to Stock

Message boards : Number crunching : Reverting to Stock
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3

AuthorMessage
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 8897 - Posted: 23 Jan 2009, 12:34:27 UTC - in response to Message 8896.  

Just keep one version only and do not accept work from other applications that are not official (released by you).

Well thats all well and good, but if I can do 5x the science for the same wattage then I want to do more for the same cost. If I can get an app that works quicker and provides the same output then I want to do more work for the same electric bill.

And then comes the famous words of the Einstein devs. Our science team will only accept results from the official app. All other results will be rejected and credit denied. I have yet to see the "electric bill" argument work on any project.


Yes, but they employed Akos and essentially rolled many / most of his improvements into the stock applications over time. I was going to suggest that Travis and Dave contact Bernd to look into the cpu detection switching apps that they are doing now to see if the same concept could be used here...


That would be awesome if it did come forwards. The benefit would be the stock client (MW Admin approved) would do the work and use the appropriate CPU instruction set detection functionality. This would cover CPUs from the older AMD and Intel ones right up to the latest (possibly not Core i7 and Phenom II).
ID: 8897 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile Kevint
Avatar

Send message
Joined: 22 Nov 07
Posts: 285
Credit: 1,076,786,368
RAC: 0
Message 8900 - Posted: 23 Jan 2009, 13:51:12 UTC
Last modified: 23 Jan 2009, 13:52:53 UTC

I found this host by accident today.

Looks like the validator was working, then not working.

http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=39517

http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=63814462

It is still using .07 and is getting valid credit. Yesterday, zero credit.

Something change ???
.
ID: 8900 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile mscharmack
Avatar

Send message
Joined: 4 Dec 07
Posts: 45
Credit: 1,257,904
RAC: 0
Message 8904 - Posted: 23 Jan 2009, 16:44:44 UTC
Last modified: 23 Jan 2009, 16:48:47 UTC

What I'm trying to say, if you have a faster application code then share it with Travis and Dave so they can incorporate it into their official code and made available to everyone, so that people are not jumping through hoops trying to install an unofficial application and creating more problems and headaches for for the administrators of this project. This is exactly what Travis and Dave have said over and over in the past and it seems to be a practical idea. They can not be expected to work out problems in some other code that was not put out by them. If you have a faster application, simple share it with them. Everyone benefits when an highly optimized code is made available.
ID: 8904 · Rating: 0 · rate: Rate + / Rate - Report as offensive
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 8905 - Posted: 23 Jan 2009, 16:48:35 UTC - in response to Message 8900.  

I found this host by accident today.

Looks like the validator was working, then not working.

http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=39517

http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=63814462

It is still using .07 and is getting valid credit. Yesterday, zero credit.

Something change ???


The way I am reading the results of this host is that up to this WU, using V0.7, it was getting credit. But for the next 3 WUs, still using V0.7, the validator has caught up with it (finally).
ID: 8905 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile Paul D. Buck

Send message
Joined: 12 Apr 08
Posts: 621
Credit: 161,934,067
RAC: 0
Message 8906 - Posted: 23 Jan 2009, 17:19:17 UTC

Just one other note.

If someone wants to only share a binary with the project for download, the project cannot accept that because of liability issues. In theory, with open source the tweakers would come up with improvements and share them back with the project which could then make them part of the baseline.

Contrast the other two projects that have optimized client applications SaH and Einstein ... EaH does all of the work in-house making changes and optimizations, beta testing them by allowing people to download the improved applications as beta, installing them, running them and as time wends on the changes become part of the baseline.

At SaH all the optimizing work is done by people off project and the sources for the changed binaries are all third party. If you want to run an improved application you must find it yourself, install it your self and maintain it your self. Improvements are not really rolled-up into the stock application.

Here, they don't really have the internal resources to do all of the experimentation that EaH does with what appears to be at least one dedicated developer that does nothing else for the project but work on making improved applications and the accumulations of years worth of improvements to the application.

So, the hope was that others would help figure out the improvements and feed information back to the project so that they could make better applications. With the information the project could then build improved binaries that we call could use.
ID: 8906 · Rating: 0 · rate: Rate + / Rate - Report as offensive
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 8910 - Posted: 23 Jan 2009, 21:28:39 UTC - in response to Message 8906.  

Just one other note.

If someone wants to only share a binary with the project for download, the project cannot accept that because of liability issues. In theory, with open source the tweakers would come up with improvements and share them back with the project which could then make them part of the baseline.


The code is now under GPL v3. So if someone is releasing binaries to the public, they must also release their source code as per the GPL.

Here, they don't really have the internal resources to do all of the experimentation that EaH does with what appears to be at least one dedicated developer that does nothing else for the project but work on making improved applications and the accumulations of years worth of improvements to the application.[/quote

Yeah, we're still a young project and it's basically me and Dave doing all the work (and Dave is only working as an undergrad researcher so thats maybe 10-15 hours a week). I also have other responsibilities as well. For what we have I think the project is coming along nicely (although a lot of you guys might disagree :P ).

[quote]So, the hope was that others would help figure out the improvements and feed information back to the project so that they could make better applications. With the information the project could then build improved binaries that we call could use.


I really hope people will be forthcoming with code improvements as they find them. The application (as far as code goes) is a million times better than the older version, so i'm not sure how much more can be squeezed out of it code-wise.
ID: 8910 · Rating: 0 · rate: Rate + / Rate - Report as offensive
jedirock
Avatar

Send message
Joined: 8 Nov 08
Posts: 178
Credit: 6,140,854
RAC: 0
Message 8915 - Posted: 23 Jan 2009, 23:06:11 UTC - in response to Message 8910.  

Just one other note.

If someone wants to only share a binary with the project for download, the project cannot accept that because of liability issues. In theory, with open source the tweakers would come up with improvements and share them back with the project which could then make them part of the baseline.


The code is now under GPL v3. So if someone is releasing binaries to the public, they must also release their source code as per the GPL.

Not quite. They are not required to release the source code with the binary, but they are required by the license to release the source code on request. Whether that be a request to a web server, an email, or even a written letter, it has to be given.
ID: 8915 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile mscharmack
Avatar

Send message
Joined: 4 Dec 07
Posts: 45
Credit: 1,257,904
RAC: 0
Message 8919 - Posted: 23 Jan 2009, 23:39:58 UTC
Last modified: 23 Jan 2009, 23:41:02 UTC

This is what happened between 0.07 and 0.13:

Computers A - App 0.07, 1.1 hrs; App 0.10, 0.13, 1.0 hrs; App 0.13, 1.2 hrs.
Computer B - 0.55 hrs; 0.51 hrs; and 0.65 hrs, respectively

Computers A - Athlon XP2600+, Athlon XP2800+ and two Athlon XP3000+
Computer B - Athlon64 X2 6000

What was changed to reduce the time App 0.07 and 0.10 and then increased the time between App 0.10 and 0.13. Are the work units longer or what?
ID: 8919 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile banditwolf
Avatar

Send message
Joined: 12 Nov 07
Posts: 2425
Credit: 524,164
RAC: 0
Message 8921 - Posted: 24 Jan 2009, 0:19:59 UTC - in response to Message 8919.  

I believe it was mentioned that the code went back to the .07 version for .13 except for the part that supposidly covers the memory leak.
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.
ID: 8921 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Misfit
Avatar

Send message
Joined: 27 Aug 07
Posts: 915
Credit: 1,503,319
RAC: 0
Message 8927 - Posted: 24 Jan 2009, 1:50:09 UTC - in response to Message 8896.  
Last modified: 24 Jan 2009, 1:50:35 UTC

Just keep one version only and do not accept work from other applications that are not official (released by you).

Well thats all well and good, but if I can do 5x the science for the same wattage then I want to do more for the same cost. If I can get an app that works quicker and provides the same output then I want to do more work for the same electric bill.

And then comes the famous words of the Einstein devs. Our science team will only accept results from the official app. All other results will be rejected and credit denied. I have yet to see the "electric bill" argument work on any project.

Yes, but they employed Akos and essentially rolled many / most of his improvements into the stock applications over time. I was going to suggest that Travis and Dave contact Bernd to look into the cpu detection switching apps that they are doing now to see if the same concept could be used here...

They had no need to employ him. He was already successful at reverse engineering and optimizing the code. In a nutshell they paid him to stop his distribution of it and make it look like they were employing him for our benefit. They must have made him an offer he couldn't refuse.
me@rescam.org
ID: 8927 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Misfit
Avatar

Send message
Joined: 27 Aug 07
Posts: 915
Credit: 1,503,319
RAC: 0
Message 8928 - Posted: 24 Jan 2009, 1:56:54 UTC - in response to Message 8906.  

Contrast the other two projects that have optimized client applications SaH and Einstein ... EaH does all of the work in-house making changes and optimizations, beta testing them by allowing people to download the improved applications as beta, installing them, running them and as time wends on the changes become part of the baseline.

Apples and oranges. Einstein is a generic stock app. The various SETI optimized apps are CPU specific. It is up to the project to create the best generic app, and up to the participants to create the best specific app.
me@rescam.org
ID: 8928 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Fred J. Verster

Send message
Joined: 22 Apr 09
Posts: 38
Credit: 27,377,932
RAC: 0
Message 40578 - Posted: 22 Jun 2010, 13:51:28 UTC - in response to Message 8928.  

AT least not a smart move, since more science projects (try to)make use use of the GPU, there is a lot to be gained, by implementing BROOK+/OpenCL/(CAL), even SETI AP works with fast GPU app.(Bêta-testing-still) but runtimes are less then a quarter and CPU times about 80 times less!


Knight Who says Ni
ID: 40578 · Rating: 0 · rate: Rate + / Rate - Report as offensive
KWSN-GMC-Peeper of the Castle ...
Avatar

Send message
Joined: 16 Dec 10
Posts: 21
Credit: 13,709,128
RAC: 0
Message 49423 - Posted: 19 Jun 2011, 14:20:29 UTC
Last modified: 19 Jun 2011, 14:27:55 UTC

This is supposed to be science, not a race to see who can accumulate the most points.
Using anything other than the project approved software is BS and you're being childish and threatening the integrity of the projects. If you don't understand WHY, you don't belong here.
In any case, all the serious science projects list as having more crunchers than needed..what's the point? All you're accomplishing is to make them issue WU to 9 crunchers instead of 3 to keep us busy. No Value Added.
I've only EVER run SETI or BOINC on my personal, home computer and i'm still well into the 98th percentile.
I don't have a spare bedroom full of frigging obsolete computers crunching score up for me, but I HAVE been active and crunching for over 12 years now with maybe 1 month's total downtime and you fanboys are NEVER going to show that kind of commitment.
;
ID: 49423 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Profile Crunch3r
Volunteer developer
Avatar

Send message
Joined: 17 Feb 08
Posts: 363
Credit: 258,227,990
RAC: 0
Message 49433 - Posted: 19 Jun 2011, 20:26:03 UTC - in response to Message 49423.  

First of all .. why do you reply to a post that's almost a year old ???

This is supposed to be science, not a race to see who can accumulate the most points.


It is science and why the hell do you think there's a system in place that counts points ??? It's a race AND it's a big part of BOINC!


Using anything other than the project approved software is BS and you're being childish and threatening the integrity of the projects. If you don't understand WHY, you don't belong here.


Actually your post is BS!... there are optimized apps out there that do SCIENCE faster than the stock apps and those apps are approved... so no need to bitch about that. Get your facts straight before posting crap like that!


Join Support science! Joinc Team BOINC United now!
ID: 49433 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Previous · 1 · 2 · 3

Message boards : Number crunching : Reverting to Stock

©2024 Astroinformatics Group