Alright, where do I go to d/l this thing? I presume it won't negatively affect any of my other projects? Will it overheat a laptop? I run einstein on one machine that is win ME and another win XP. Thanks.
Founder of BOINC group, Objectivists, a group of philosophically minded rational data crunchers.
Alright, where do I go to d/l this thing? I presume it won't negatively affect any of my other projects? Will it overheat a laptop? I run einstein on one machine that is win ME and another win XP. Thanks.
You can download the optimized apps here. Just check with something like cpu-z what instructions your computer supports.
Lets try to keep this a thread dedicated to Akos
There are 10^11 stars in the galaxy. That used to be a huge number. But it's only a hundred billion. It's less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers. - Richard Feynman
Thank you Akos for your brilliant work - the speed improvements are breath-taking. And thanks a lot for speeding up all - even very old - x86 CPUs with your optimized clients.
Chris397
Alright, where do I go to d/l this thing? I presume it won't negatively affect any of my other projects? Will it overheat a laptop? I run einstein on one machine that is win ME and another win XP. Thanks.
You can download the optimized apps here. Just check with something like cpu-z what instructions your computer supports.
Lets try to keep this a thread dedicated to Akos
I think it's a bit ridiculous that you want this thread to exist so people can thank him for creating his optimized work but discourage people from asking how they should actually go about using it. Especially since this thread is/may be the first place some readers have ever heard of it.
Founder of BOINC group, Objectivists, a group of philosophically minded rational data crunchers.
I think it's a bit ridiculous that you want this thread to exist so people can thank him for creating his optimized work but discourage people from asking how they should actually go about using it. Especially since this thread is/may be the first place some readers have ever heard of it.
I am sorry if that was the tone that you got from my post. I think it is great that you are asking questions and want to optimize your computer. I was in no way trying to discourage you or any others from asking these questions, I even provided a link to where you can find the answers. I was simply trying to state that this thread is not necessarily the place to post you question, start a new one, or take a look around and see if someone has already asked/answered your question in another thread. Akos has done a lot for the E@H community, and I didn't want this thread to get too side-tracked (as they tend to, and I admit to being guilty of that at the moment) from users sharing their thanks.
There are 10^11 stars in the galaxy. That used to be a huge number. But it's only a hundred billion. It's less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers. - Richard Feynman
To Akos...what a guy..to put your time and talent to making this number crunching so "much" more interesting. I have been updating with each successive application you have provided...
I too wish to express my appreciation to Akos. It's simply breathtaking what my old crusty machines are capable of now. I've set my systems to focus on crunching just this project now until it's finished, so your efforts have not only made these systems much more efficiently utilized, but had the side effect of stoking up greater interest on my part on this project. Here's hoping that the visibility you get in this virtual world will result in rewards in the real one.
Akos, I've never imagined crunching times that we're seeing now.
Your programming skills are very impressive, too bad Windows
programmers aren't more like you!
Thanks and Well Done!
There're two different factors at work behind the speedup. The quality of the algorithm, and how efficiently it's translated into machine instructions. The later is generally a function of the compiler in converting from a high level design into assembly language. This can be done by either taking a c/c++/fortran/etc program which is easily and quicly writable and compiling it, or writing the asm by hand. The traditional number for c vs asm development time is 10:1. As compilers get more sophisticated the ammount of knowledge of exactly how the processor works needed to gain better performance grows increasingly high. Also, the vast majority of code in an app is seldom executed and hand tweaking would show little or no benefit. This, combined with the 10:1 devtime is why modern apps are almost never written in asm anymore.
The first factor, quality of algorithm is as much a factor of time/effort invested as anything else. And here in lies the rub. LIGO is a hardware project, not a software one. And as every EE/ME/etc knows, the hardware is the hard part where all the time/effort/money needs to be spent. Software on the other hand is quick/easy/cheap to develop, and can be given a tiny slice of the budget.
If Akos is reading this, I'd be interested in knowing what the breakdown on the speedup is between better algorithms (I know his 1st version was lagely this), and better implementation of existing ones (his most recent SSE/x87 fpu interleaving is this). And also how much of the code he's actually rewritten in a meaningful manor. I could do a binary diff, but anywhere replacement code was a different length than the original it would result in changed offsets throughout the app that would generate lots of false positives.
Alright, where do I go to d/l
)
Alright, where do I go to d/l this thing? I presume it won't negatively affect any of my other projects? Will it overheat a laptop? I run einstein on one machine that is win ME and another win XP. Thanks.
Founder of BOINC group, Objectivists, a group of philosophically minded rational data crunchers.
RE: Alright, where do I go
)
You can download the optimized apps here. Just check with something like cpu-z what instructions your computer supports.
Lets try to keep this a thread dedicated to Akos
There are 10^11 stars in the galaxy. That used to be a huge number. But it's only a hundred billion. It's less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers. - Richard Feynman
Thank you Akos for your
)
Thank you Akos for your brilliant work - the speed improvements are breath-taking. And thanks a lot for speeding up all - even very old - x86 CPUs with your optimized clients.
Chris397
respect dude but remember, we
)
respect dude
but remember, we all need sleep eventually ;)
RE: RE: Alright, where do
)
I think it's a bit ridiculous that you want this thread to exist so people can thank him for creating his optimized work but discourage people from asking how they should actually go about using it. Especially since this thread is/may be the first place some readers have ever heard of it.
Founder of BOINC group, Objectivists, a group of philosophically minded rational data crunchers.
RE: I think it's a bit
)
I am sorry if that was the tone that you got from my post. I think it is great that you are asking questions and want to optimize your computer. I was in no way trying to discourage you or any others from asking these questions, I even provided a link to where you can find the answers. I was simply trying to state that this thread is not necessarily the place to post you question, start a new one, or take a look around and see if someone has already asked/answered your question in another thread. Akos has done a lot for the E@H community, and I didn't want this thread to get too side-tracked (as they tend to, and I admit to being guilty of that at the moment) from users sharing their thanks.
There are 10^11 stars in the galaxy. That used to be a huge number. But it's only a hundred billion. It's less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers. - Richard Feynman
To Akos...what a guy..to put
)
To Akos...what a guy..to put your time and talent to making this number crunching so "much" more interesting. I have been updating with each successive application you have provided...
Your "talent" is appreciated!!
I too wish to express my
)
I too wish to express my appreciation to Akos. It's simply breathtaking what my old crusty machines are capable of now. I've set my systems to focus on crunching just this project now until it's finished, so your efforts have not only made these systems much more efficiently utilized, but had the side effect of stoking up greater interest on my part on this project. Here's hoping that the visibility you get in this virtual world will result in rewards in the real one.
RE: Akos, I've never
)
There're two different factors at work behind the speedup. The quality of the algorithm, and how efficiently it's translated into machine instructions. The later is generally a function of the compiler in converting from a high level design into assembly language. This can be done by either taking a c/c++/fortran/etc program which is easily and quicly writable and compiling it, or writing the asm by hand. The traditional number for c vs asm development time is 10:1. As compilers get more sophisticated the ammount of knowledge of exactly how the processor works needed to gain better performance grows increasingly high. Also, the vast majority of code in an app is seldom executed and hand tweaking would show little or no benefit. This, combined with the 10:1 devtime is why modern apps are almost never written in asm anymore.
The first factor, quality of algorithm is as much a factor of time/effort invested as anything else. And here in lies the rub. LIGO is a hardware project, not a software one. And as every EE/ME/etc knows, the hardware is the hard part where all the time/effort/money needs to be spent. Software on the other hand is quick/easy/cheap to develop, and can be given a tiny slice of the budget.
If Akos is reading this, I'd be interested in knowing what the breakdown on the speedup is between better algorithms (I know his 1st version was lagely this), and better implementation of existing ones (his most recent SSE/x87 fpu interleaving is this). And also how much of the code he's actually rewritten in a meaningful manor. I could do a binary diff, but anywhere replacement code was a different length than the original it would result in changed offsets throughout the app that would generate lots of false positives.
Akos hits The News! See
)
Akos hits The News!
See the article in New Scientist.
Well done, good stuff, happy hackings,
And happy crunchin',
Regards,
Martin
See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)