I want to make optimized clients.....

Tetsuji Maverick Rai
Tetsuji Maverick Rai
Joined: 11 Apr 05
Posts: 23
Credit: 3,658,667
RAC: 0
Topic 189510

I've made optimized clients for SETI@home and SETI@home beta test because the sources are available.

(see this one for SETI@home and this for seti beta.)

So if the source were available on Einstein, I would have a chance to make optimized clients.......

[B@H] Ray
[B@H] Ray
Joined: 4 Jun 05
Posts: 621
Credit: 49,583
RAC: 0

I want to make optimized clients.....

Please Please Mr. Einstein make your source code available to Tetsuji for making a cpu optomized client. Trtsuji did great with the seti ones.

Ray


Try the Pizza@Home project, good crunching.

Tetsuji Maverick Rai
Tetsuji Maverick Rai
Joined: 11 Apr 05
Posts: 23
Credit: 3,658,667
RAC: 0

RE: I've made optimized

Quote:

I've made optimized clients for SETI@home and SETI@home beta test because the sources are available.

(see this one for SETI@home and this for seti beta.)

So if the source were available on Einstein, I would have a chance to make optimized clients.......

but forget it :)....I am comfortable with SETI@home and the beta test :) Astropulse may be waiting..

Revolution
Revolution
Joined: 17 Jun 05
Posts: 11
Credit: 14,349
RAC: 0

Yeah - Lets all have our own

Yeah - Lets all have our own version of the code!
Maybe we can give them 100 different results for the same data!

Ocean Archer
Ocean Archer
Joined: 18 Jan 05
Posts: 92
Credit: 368,644
RAC: 0

RE: Yeah - Lets all have

Message 13951 in response to message 13950

Quote:
Yeah - Lets all have our own version of the code!
Maybe we can give them 100 different results for the same data!

Revolution --

Please do not belittle the contribution made by Tetsuji (or any other cruncher for that matter). We ALL are volunteers with a common cause -- to provide distributed computing power to worthwhile projects. The optimized code for SETI@home and SETI beta are excellent examples of work and are well received by the project heads. It would be nice if we could get one for E@H too, but that's up to the Einstein heads.


If I've lived this long - I gotta be that old!

Digitalis
Digitalis
Joined: 9 Feb 05
Posts: 18
Credit: 90,764
RAC: 0

RE: RE: Yeah - Lets all

Message 13952 in response to message 13951

Quote:
Quote:
Yeah - Lets all have our own version of the code!
Maybe we can give them 100 different results for the same data!

Revolution --

Please do not belittle the contribution made by Tetsuji (or any other cruncher for that matter). We ALL are volunteers with a common cause -- to provide distributed computing power to worthwhile projects. The optimized code for SETI@home and SETI beta are excellent examples of work and are well received by the project heads. It would be nice if we could get one for E@H too, but that's up to the Einstein heads.


I too can testify to the excellent work of Tetsuji in optimising the Seti and SetiBeta apps and the extremely high regard he is held in by that community. In particular his inclusion of SSE2 instructions has the desireable effect of halving the processing time of a Seti wu on a P4 cpu with absolutely no degradation of the quality of results.
Since the Einstein app is not open source only the developers can say if it is capable of being optimised, not all project apps are e.g. LHCs Sixtrack. I look forward to hearing their views on this subject.

[B@H] Ray
[B@H] Ray
Joined: 4 Jun 05
Posts: 621
Credit: 49,583
RAC: 0

RE: I too can testify to

Message 13953 in response to message 13952

Quote:
I too can testify to the excellent work of Tetsuji in optimising the Seti and SetiBeta apps and the extremely high regard he is held in by that community. In particular his inclusion of SSE2 instructions has the desireable effect of halving the processing time of a Seti wu on a P4 cpu with absolutely no degradation of the quality of results.
Since the Einstein app is not open source only the developers can say if it is capable of being optimised, not all project apps are e.g. LHCs Sixtrack. I look forward to hearing their views on this subject.

Yes his work on the seti program for SSE2 is great, I use it and really like the way it works. Besides running faster it is also better at letting other resource hogging programs run ahead of it so you don't even notice that seti is running.

With the source not available maby the prople at Einstein will have optomixed clients for SSE2, SSE3 and other instructin sets. It would only be towards there advantage to have them run faster and better on computer systems. Maby more would stay with Einstein longer.

Thank you Tetsuji for your great work on the seti clients.

Ray


Try the Pizza@Home project, good crunching.

Shaktai
Shaktai
Joined: 8 Nov 04
Posts: 183
Credit: 426,451
RAC: 0

In the beta test thread for

In the beta test thread for Windows (Cafe Einstein)Bernard posted the following:

Quote:

I once compiled the code with SSE optimization and found it to be about 1% faster than without, so there's not much in the normal code that the compiler found to be suitable for SSE.

One _might_ get further porting the vector code that was (re-)written for the AltiVec to SSE (actually SS2 as I doubt that SSE offers enough precision) but this would need more work again and would end up with yet another code version to be maintained. We are aiming for the "auto-vectorize" feature of the new gcc-4, which should give us generic code as well as code for multiple vector units from the same source code, but it will will also take some more time until it is mature enough to be of use for us.

BM

The Einstein app code cannot be released publically, and it would be difficult for the project to support multiple versions. However Bernard might be open to suggestions for optimizations flags that could be used, and with what compiler.

Revolution
Revolution
Joined: 17 Jun 05
Posts: 11
Credit: 14,349
RAC: 0

I have no idea who Tetsuji

I have no idea who Tetsuji is.
For any project such as this to succeed all clients *should* be returning Exactly the same data results.

Open source code is a bad idea but I agree that certain compile codes can improve calculation performance. This IMO should be up to the project managers.

AnRM
AnRM
Joined: 9 Feb 05
Posts: 213
Credit: 4,346,941
RAC: 0

RE: I have no idea who

Message 13956 in response to message 13955

Quote:
I have no idea who Tetsuji is.
Open source code is a bad idea but I agree that certain compile codes can improve calculation performance. This IMO should be up to the project managers.


I agree totally......it would be nice to have an updated Linux code though. It seems to be an lonely orphan.....

Enno Ruijters
Enno Ruijters
Joined: 9 Feb 05
Posts: 2
Credit: 743,521
RAC: 84

Revolution: The results

Revolution:

The results returned by the optimized clients are the same as the results made by the official client. The only difference would be because of some roundoff error, and those errors are already causing minor differences on different platforms/OS's.

Besides, if the clients were returning bad results, then the users of those clients would not get any credit and the client would not be used anyway.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.