Since we cant test Akos codes on the live project, we would need to do it on a beta server. Ether the project could set one up, or someone trusted outside the project could run it, perhaps Pirates could be convinced.
The results should preferable be small, so using one of the smaller datasets in the lower frequencies maybe (short result, small download). If the project supports this idea, we could use the canonical results for this dataset from the main project to validate the results against. The initial replication should be one, and the validator would track the differences between the test result and the canonical result.
Then a stable version is found (with sufficient improvement) Bernard could evaluate the application and if he is happy with it, he post it as an official Beta test application on the main project.
Then you're really interested in a subject, there is no way to avoid it. You have to read the Manual.
Then a stable version is found (with sufficient improvement) Bernard could evaluate the application and if he is happy with it, he post it as an official Beta test application on the main project.
Why as a beta? After testing throuroughly in a beta-project it could well go as an official client to be used optionally.
Or, like the folks at Simap, the Einsteins could make it the client that is officionally loaded by the Boinc-client with a client for older, non-SSE-CPUs to be downloaded and implemented manually.
In the end that would decrease the time for going through S5 significantly, freeing capacitiy for S6 or, as good, just for other projects.
Yes, that would be wonderful if that would work that way, but the official application has to be able run on all hosts. The official application can only differentiate between SSE and non SSE hosts, and would therefore have to be optimized for booth SSE and non SSE hosts. An SSE3 application would therefore have to be an official Beta application.
Edit: the Einstein application is smart enough to detect if the host is capable of SSE or not. Simaps application isn’t.
Then you're really interested in a subject, there is no way to avoid it. You have to read the Manual.
Why isn't it possible to create a SMALL beta Project for each optimisation?
Perhaps, a group of 50 Crunchers with different CPUs (which are matching with the patch) could check the Version ... 10 with the stock-application, the others with the patched one ...
After that, someone needs to check the results, in the best case, they would be absolutly exact.
You profit from these better applications too!
with kind regards ...
dark-enforcer
sorry for my bad english
Could count me in too :)
I'm curious. Theres still a link to the S4 beta area on the front page. Firstly thats out of date except maybe for the FreeBSD x86's 4.09 beta. Think someone brought that up already but also why not use that towards this? Bascially works on same pricipal was happening here-go there and download the beta app but akosf's wasn't official. The project leaders would have to have a look at the code akosf was putting forward and make a beta app out of it and call it an official beta. I get the impression this was what they were aiming at with the report as 4.10 idea. Make it 4.10 meaning running as beta.
I think thats probly what they are looking at doing anyway by posssibly putting some of his work in the standard app.
My point, there is a beta area which E@H uses and has no quarrel with but the apps have to be from them or the scientific comunity might get iffy about the results. Currently there are no official beta apps to test except for FreeBSD x86. When they have others ready i'll gradly join in putting it throught its paces. The offiaciality of it all is the problem for E@H.
Sorry about the bad grammer (and probly spelling too) :)
Yes, that would be wonderful if that would work that way, but the official application has to be able run on all hosts. The official application can only differentiate between SSE and non SSE hosts, and would therefore have to be optimized for booth SSE and non SSE hosts. An SSE3 application would therefore have to be an official Beta application.
Edit: the Einstein application is smart enough to detect if the host is capable of SSE or not. Simaps application isn’t.
This is being worked on at the moment by Rom Walton. Upcoming releases of the BOINC Client will be able to detect and report your processor's instruction set, making it possible for BOINC projects to distribute processor-specific applications that make use of SSE2, SSE3, and so on:
Nicely stated, Mike. I am a litlle put off though, since I was crafting a very similar post on an Office document when I read yours. I am frustrated to say that yours is better, so I'll just say that I agree with your sentiment.
Thanks, but don't be shy, I'm not competing! :-)
Quote:
True science is done correctly only if it is done deliberately and slowly. There can be no shortcuts to true knowledge and certainly none at all to true understanding and enlightenment.
Optimizations are to be welcomed with open arms but only if they produce results that are unimpugnable by the scientific community at large.
Oh so true!
Quote:
We love you akosf, and know that you hold the integrity of the project to heart. Your efforts, and those equal efforts of the others at the core of this project, towards improving the efficiency of the process of discovery are commendable.
I love seeing lots of work units flying out of my mini farm, but knowing that my results are truly valid is the only important consideration. None the less, I'd love to see optimizations integrated ASAP.
I know I'm only one of many that will gladly volunteer cycles to test apps and clients to forward the process.
Best wishes to all the community,
Jim
Akos - I do hope this kerfuffle doesn't put you off at all! :-)
Cheers, Mike.
( edit ) kerfuffle = storm in a tea cup.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
Akos - I do hope this kerfuffle doesn't put you off at all! :-)
Cheers, Mike.
( edit ) kerfuffle = storm in a tea cup.
I like this project. I can always learn new words and expressions. :-)
But i think my english will stay very poor. I'm not a language genius. :-)
I think the results of this project will be very useful in our everydays, so i hope we will finish with all WUs as fast as possible. I'm a bit sad that some people left this project and i also have to detach my computers, but the community is really good and i know the people at the Einstein@Home also try to do their best.
RE: RE: Feel free to
)
That's what i tought ;)
Anyway here's a banner for SETI.Germany....
Feel free to use it ;)
RE: Feel free to use it
)
cool, big THX !!!
RE: RE: Feel free to use
)
No problem my friend ;)
Since we cant test Akos codes
)
Since we cant test Akos codes on the live project, we would need to do it on a beta server. Ether the project could set one up, or someone trusted outside the project could run it, perhaps Pirates could be convinced.
The results should preferable be small, so using one of the smaller datasets in the lower frequencies maybe (short result, small download). If the project supports this idea, we could use the canonical results for this dataset from the main project to validate the results against. The initial replication should be one, and the validator would track the differences between the test result and the canonical result.
Then a stable version is found (with sufficient improvement) Bernard could evaluate the application and if he is happy with it, he post it as an official Beta test application on the main project.
Then you're really interested in a subject, there is no way to avoid it. You have to read the Manual.
RE: Then a stable version
)
Why as a beta? After testing throuroughly in a beta-project it could well go as an official client to be used optionally.
Or, like the folks at Simap, the Einsteins could make it the client that is officionally loaded by the Boinc-client with a client for older, non-SSE-CPUs to be downloaded and implemented manually.
In the end that would decrease the time for going through S5 significantly, freeing capacitiy for S6 or, as good, just for other projects.
Greetings groundhog
Yes, that would be wonderful
)
Yes, that would be wonderful if that would work that way, but the official application has to be able run on all hosts. The official application can only differentiate between SSE and non SSE hosts, and would therefore have to be optimized for booth SSE and non SSE hosts. An SSE3 application would therefore have to be an official Beta application.
Edit: the Einstein application is smart enough to detect if the host is capable of SSE or not. Simaps application isn’t.
Then you're really interested in a subject, there is no way to avoid it. You have to read the Manual.
RE: Why isn't it possible
)
Could count me in too :)
I'm curious. Theres still a link to the S4 beta area on the front page. Firstly thats out of date except maybe for the FreeBSD x86's 4.09 beta. Think someone brought that up already but also why not use that towards this? Bascially works on same pricipal was happening here-go there and download the beta app but akosf's wasn't official. The project leaders would have to have a look at the code akosf was putting forward and make a beta app out of it and call it an official beta. I get the impression this was what they were aiming at with the report as 4.10 idea. Make it 4.10 meaning running as beta.
I think thats probly what they are looking at doing anyway by posssibly putting some of his work in the standard app.
My point, there is a beta area which E@H uses and has no quarrel with but the apps have to be from them or the scientific comunity might get iffy about the results. Currently there are no official beta apps to test except for FreeBSD x86. When they have others ready i'll gradly join in putting it throught its paces. The offiaciality of it all is the problem for E@H.
Sorry about the bad grammer (and probly spelling too) :)
RE: Yes, that would be
)
This is being worked on at the moment by Rom Walton. Upcoming releases of the BOINC Client will be able to detect and report your processor's instruction set, making it possible for BOINC projects to distribute processor-specific applications that make use of SSE2, SSE3, and so on:
http://setiathome.berkeley.edu/forum_thread.php?id=31879#339634
Dig
RE: Nicely stated, Mike. I
)
Thanks, but don't be shy, I'm not competing! :-)
Oh so true!
Akos - I do hope this kerfuffle doesn't put you off at all! :-)
Cheers, Mike.
( edit ) kerfuffle = storm in a tea cup.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
RE: Akos - I do hope this
)
I like this project. I can always learn new words and expressions. :-)
But i think my english will stay very poor. I'm not a language genius. :-)
I think the results of this project will be very useful in our everydays, so i hope we will finish with all WUs as fast as possible. I'm a bit sad that some people left this project and i also have to detach my computers, but the community is really good and i know the people at the Einstein@Home also try to do their best.
Let's crunch!
More gravity?
Akos