Optomized S5 SSE3

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9352143
RAC: 0

RE: That sounds like

Message 39440 in response to message 39435

Quote:

That sounds like complete baloney to me.

Please convince me that the Einstein@Home distributed computing project is more than a filter to find promising candidates. There is no way the E@H scientists would publish any findings without recrunching any and all signals that resemble the detection of gravitational waves under tightly controlled circumstances.

30%+ is a huge gain to throw away with that sort of bogus argument.

In Bruce's electricity saving terms: $10K per day in electicity -> some number crunching -> at least 20% electricity savings with optimized apps -> at least $2k per day in savings foregone without current level of optimization -> somewhere between $750K and $1M in energy wasted and paid for by all crunchers over the course of S5.

With all due respect and no "flame" intended:

I haven't personally read the study protocol, have you?

Since Akos inferred he received word from Bernd or Dr. Allen about a potential problem, who are you to draw conclusions like that?

As far as the power costs to run EAH goes, if you have to worry about that then you probably shouldn't be running DC projects at all.

Alinator

Dotshead
Dotshead
Joined: 12 Jan 06
Posts: 13
Credit: 69993
RAC: 0

RE: IMHO: I think the

Message 39441 in response to message 39436

Quote:
IMHO: I think the reactions here are overdone. We should wait for an "official" statement from Bruce, et. al. Putting some controls in the WU distribution process is the prudent thing to do (and its need should have been anticipated by the developers when they went to the 2 result quorum). But, give them some time to explain things to us. My guess is Akos' work will be allowed back in sometime in the future, when a better process is in place.

How long should we wait to get an "official" statement from Bruce? How long did it take time since Akos optimations were recognized from the "official" E@H-Team for the S4 run?

Dots

Stick
Stick
Joined: 24 Feb 05
Posts: 790
Credit: 33144875
RAC: 979

RE: RE: IMHO: I think the

Message 39442 in response to message 39441

Quote:
Quote:
IMHO: I think the reactions here are overdone. We should wait for an "official" statement from Bruce, et. al. Putting some controls in the WU distribution process is the prudent thing to do (and its need should have been anticipated by the developers when they went to the 2 result quorum). But, give them some time to explain things to us. My guess is Akos' work will be allowed back in sometime in the future, when a better process is in place.

How long should we wait to get an "official" statement from Bruce? How long did it take time since Akos optimations were recognized from the "official" E@H-Team for the S4 run?

Dots

He needs to say something ASAP, else he will lose a lot of crunchers (not to mention credibility). But we should withhold judgement until he does.

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9352143
RAC: 0

RE: How long should we

Message 39443 in response to message 39441

Quote:


How long should we wait to get an "official" statement from Bruce? How long did it take time since Akos optimations were recognized from the "official" E@H-Team for the S4 run?

Dots

If I understand you here correctly, Dr. Allen responded to the appearance of the S4 apps within a week or so of their arrival.

IIRC, they were never "officially" sanctioned as a replacement for the stock Albert app.

Alinator

Dronak
Dronak
Joined: 21 Mar 05
Posts: 28
Credit: 10402879
RAC: 0

Very disappointing. Granted,

Very disappointing. Granted, not every patch has been good with validated results. But that's part of testing, and akosf has managed to make some great gains in speed by optimizing the application code. As already noted, it was about 30% faster. I do wish this was handled better. I really hope that akosf's improvements are not lost, and the people behind the project either (1) start allowing optimized applications or (2) incorporate those optimizations into the stock application.

I'm not going to quit the project, but I will say that this has caused Einstein@home to lose favor with me. I used to be able to finish a work unit in 1 hour before with optimized applications, but now long work units will take 12+ hours to complete and I'll be lucky to return even 1 work unit a day. I've gone back to using the stock application as requested because it's only right to do so. I certainly don't want to mess up the project's work. However, I've cut back the resource share it gets, and will allow my computer to spend more time on other projects instead of this one. If the situation improves, I may put more focus here again, but I won't hold my breath. Very disappointing.

MRAO
MRAO
Joined: 7 May 05
Posts: 33
Credit: 15770746
RAC: 0

BTW, if anyone feels like a

BTW, if anyone feels like a change for a day or two until this settles down, SZTAKI has just announced that its new WUs will be available there tomorrow. Don't trample each other in the rush please. :-) Mike

PS - Akos got a 50 times improvement there previously, not just getting on for 50%! What is the Hungarian for genius??

Akos Fekete
Akos Fekete
Joined: 13 Nov 05
Posts: 561
Credit: 4527270
RAC: 0

RE: Where is the problem

Message 39446 in response to message 39433

Quote:
Where is the problem and where do you cause problems due to your speed optimations? Or is the problem that you increase the speed unoffically?

The problem is not the speed.
The problem is that these apps are unofficials.

Dotshead
Dotshead
Joined: 12 Jan 06
Posts: 13
Credit: 69993
RAC: 0

RE: RE: How long should

Message 39447 in response to message 39443

Quote:
Quote:


How long should we wait to get an "official" statement from Bruce? How long did it take time since Akos optimations were recognized from the "official" E@H-Team for the S4 run?

Dots

If I understand you here correctly, Dr. Allen responded to the appearance of the S4 apps within a week or so of their arrival.

IIRC, they were never "officially" sanctioned as a replacement for the stock Albert app.

Alinator

Thank you for blaming Akos. Akos did a very good job for S4 instead of the personal which was paid for. Akos should be an "official" member of the team.

I think everybody saved a lot of energy using akos optimations on S4.

Dots

Barrie
Barrie
Joined: 23 Mar 05
Posts: 219
Credit: 21449106
RAC: 0

RE: RE: Where is the

Message 39448 in response to message 39446

Quote:
Quote:
Where is the problem and where do you cause problems due to your speed optimations? Or is the problem that you increase the speed unoffically?
The problem is not the speed.
The problem is that these apps are unofficials.

Hi Akos

Why did Bernd not tell you this instead of asking for a version number change?

You could restore some goodwill for me by answering these question, although they are academic now:

Was there ever any intention that version 4.10 would be subject to more lax validation? I did not believe this, but some did.

Was S5T0713 functionally the same as S5T0712? Some people think it was, and it produced a 0712 message. There is plenty of evidence it was not the same. Please clarify, before I submit myself as a voluntary mental patient!

Thanks.

Dead men don't get the baby washed. HTH

Dotshead
Dotshead
Joined: 12 Jan 06
Posts: 13
Credit: 69993
RAC: 0

RE: RE: Where is the

Message 39449 in response to message 39446

Quote:
Quote:
Where is the problem and where do you cause problems due to your speed optimations? Or is the problem that you increase the speed unoffically?
The problem is not the speed.
The problem is that these apps are unofficials.

rotfl

Yes this is what I expected.

Thank you for be honest.

Best regards from Mettmann (Germany)

Dots

Comment viewing options

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