Why did Bernd not tell you this instead of asking for a version number change?
They are confused. They would like to the best for the science, and the good and reliable results are the first. Version number changing was a temporary idea.
Why did Bernd not tell you this instead of asking for a version number change?
They are confused. They would like to the best for the science, and the good and reliable results are the first. Version number changing was a temporary idea.
They are confused. Yes it's a very good description for the E@H-Team. It seems that conversation is leaking within contributors to the clients.
This project is a terrific one. There is no way I would have crunched nearly one million Einstein credits, if I didn't share your enthusiasm for the project ... and we are all just pawns in the crunching universe compared to the project scientists whose whole careers are tied up in this project.
What I care most about from my outside contributors perspective is that I want to see a reasonable degree analytical thinking applied to the optimal running of the E@H distributed computing project. It would seem that: "identical results" are just that "identical". Code can be analyzed and determined to yield identical results. Code obviously isn't identical across platforms anyway: certainly not across different FPU implementations by AMD, Intel, IBM and all the others. Compilers and math libraries are not identical ... nor are they perfect ... ask the LHC people at CERN about that one. About a year ago they had a situation where different platforms always returned different results.
Let's waste some energy/money for a while. I just hope that you, Akos, can get back to being a contributor as quickly as possible.
Quote:
Quote:
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.
You should know that always the base researches need the most money and they produces the worst efficiency. But they gives the best things for the mankind.
I think most people of reasonable intelligence realized from the start that we were merely helping you to 'test' new optimizations, and that they could be pulled at any time. We also know that you will likely continue your work on the optimizations, but possibly in a more structured environment. I myself had to cringe a few times when I saw how many results were being trashed by people's carelessness. The optimized applications should be considered 'bonuses' by us, not demanded from the project.
At any rate, keep up the good work. I hope the project team didn't smack your peepee too hard over this.
Perhaps the S4 results can be legitimised by the certainty that every WU was processed at least once by an official app. and corroborating results came from unofficial apps. Has the database been analysed for this?
So we've got to stop using the patches because the work they do wont be accepted be the scientific comunity. Thats annoying to say the least. Your help was nocking at least 3 hours of each wu.
I got to ask then. Does this mean all that work done with your Unofficial optimized apps from akosf on the S4 project also 'invalid'. Seems exactle the same situation to me. You just optimized to app they put out. Then the question leads to if the standard S5 app can be considered 'valid' as most of that is based around the old albert app (bit of beta) and the improvements you(Akosf) made.
Seems silly to me. The old app you didn't have to meet as near strict validation it seems. So to me it seems the work you've done on S5 is more 'valid' than before.
I got an easy answer for all this MAKE AKOSF AN HONORARY OFFICIAL MEMBER OF THE PROJECT AND MAKE HIS WORK OFFICIAL :) you've already put his work into the 'official app' Not to imply the project runners dont do their jobs already but the project will be at a great loss to have to dissmis such good work. Over 30% improvement with no impact to validation is not something to be just neglected.
Heres hopeing Akosf's work will again make to to the official app. On the provision you DONT make the WU any longer again. PLEASE. Anyway i'm back on the standard app that i'm having to run again. Here come the times over 11hours again. Sigh. oh well was good whilst it lasted.
@Akosf- Again. Thank you so much for your efforts, dedication and incredible work. I'm very sorry things have gone this way. You can however still rest in the knowledge that you were still able to squeeze over 35% out of an already optimized app. simply Amazing.
Seems project after project is making it very dificult for the optimizers. Keep it up and you'll chase them away.
RE: RE: Why did Bernd not
)
ROTFL! Thanks again.
Dead men don't get the baby washed. HTH
akosf, are the opt. S4
)
akosf,
are the opt. S4 clients official ?
Hi! RE: Hi akosf! Are
)
Hi!
As far as i know there were only officially optimized clients in beta status.
RE: RE: Why did Bernd not
)
They are confused. Yes it's a very good description for the E@H-Team. It seems that conversation is leaking within contributors to the clients.
Dots
This project is a terrific
)
This project is a terrific one. There is no way I would have crunched nearly one million Einstein credits, if I didn't share your enthusiasm for the project ... and we are all just pawns in the crunching universe compared to the project scientists whose whole careers are tied up in this project.
What I care most about from my outside contributors perspective is that I want to see a reasonable degree analytical thinking applied to the optimal running of the E@H distributed computing project. It would seem that: "identical results" are just that "identical". Code can be analyzed and determined to yield identical results. Code obviously isn't identical across platforms anyway: certainly not across different FPU implementations by AMD, Intel, IBM and all the others. Compilers and math libraries are not identical ... nor are they perfect ... ask the LHC people at CERN about that one. About a year ago they had a situation where different platforms always returned different results.
Let's waste some energy/money for a while. I just hope that you, Akos, can get back to being a contributor as quickly as possible.
RE: Hi!RE: Hi akosf! Are
)
and hundreds or thousands peoples use it... ;)
officially optimized clients
)
officially optimized clients in beta status.
official = beta
What now???
Stay tuned and keep crunching
Akos, we do thank you for
)
Akos, we do thank you for your efforts.
I think most people of reasonable intelligence realized from the start that we were merely helping you to 'test' new optimizations, and that they could be pulled at any time. We also know that you will likely continue your work on the optimizations, but possibly in a more structured environment. I myself had to cringe a few times when I saw how many results were being trashed by people's carelessness. The optimized applications should be considered 'bonuses' by us, not demanded from the project.
At any rate, keep up the good work. I hope the project team didn't smack your peepee too hard over this.
Dig
Perhaps the S4 results can be
)
Perhaps the S4 results can be legitimised by the certainty that every WU was processed at least once by an official app. and corroborating results came from unofficial apps. Has the database been analysed for this?
Dead men don't get the baby washed. HTH
So we've got to stop using
)
So we've got to stop using the patches because the work they do wont be accepted be the scientific comunity. Thats annoying to say the least. Your help was nocking at least 3 hours of each wu.
I got to ask then. Does this mean all that work done with your Unofficial optimized apps from akosf on the S4 project also 'invalid'. Seems exactle the same situation to me. You just optimized to app they put out. Then the question leads to if the standard S5 app can be considered 'valid' as most of that is based around the old albert app (bit of beta) and the improvements you(Akosf) made.
Seems silly to me. The old app you didn't have to meet as near strict validation it seems. So to me it seems the work you've done on S5 is more 'valid' than before.
I got an easy answer for all this
MAKE AKOSF AN HONORARY OFFICIAL MEMBER OF THE PROJECT AND MAKE HIS WORK OFFICIAL :) you've already put his work into the 'official app' Not to imply the project runners dont do their jobs already but the project will be at a great loss to have to dissmis such good work. Over 30% improvement with no impact to validation is not something to be just neglected.
Heres hopeing Akosf's work will again make to to the official app. On the provision you DONT make the WU any longer again. PLEASE. Anyway i'm back on the standard app that i'm having to run again. Here come the times over 11hours again. Sigh. oh well was good whilst it lasted.
@Akosf- Again. Thank you so much for your efforts, dedication and incredible work. I'm very sorry things have gone this way. You can however still rest in the knowledge that you were still able to squeeze over 35% out of an already optimized app. simply Amazing.
Seems project after project is making it very dificult for the optimizers. Keep it up and you'll chase them away.