Docking@home display Wu and use GPU ? like Folding@home ?
@+
*_*
We are not supporting GPUs yet. But a student is starting a semester project to move our MD code for GPU to BOINC. If successful, we will have a D@H for GPU in summer.
Stay tuned!
____________
If you are interested in working on Docking@Home in a great group at UDel, contact me at 'taufer at acm dot org'!
We are not supporting GPUs yet. But a student is starting a semester project to move our MD code for GPU to BOINC. If successful, we will have a D@H for GPU in summer.
Stay tuned!
Great news!!
But, what type of client? CUDA? OpenCL? And it runs in double or single precision??
ID:
6264 | Rating: 0
| rate:
/
Michela
Forum moderator
Project administrator
Project developer
Project tester
Project scientist
We are not supporting GPUs yet. But a student is starting a semester project to move our MD code for GPU to BOINC. If successful, we will have a D@H for GPU in summer.
Stay tuned!
Great news!!
But, what type of client? CUDA? OpenCL? And it runs in double or single precision??
So far we have been working with CUDA and C (http://gcl.cis.udel.edu/projects/fenzi/index.php. The code can be compiled in both single and double precision but testing shows that for the simulations we are running single precision is sufficient.
Michela
____________
If you are interested in working on Docking@Home in a great group at UDel, contact me at 'taufer at acm dot org'!
So far we have been working with CUDA and C (http://gcl.cis.udel.edu/projects/fenzi/index.php. The code can be compiled in both single and double precision but testing shows that for the simulations we are running single precision is sufficient.
Michela
Thanks for answer!!
If you need beta testers, make a thread on forum :-)
Sorry, but if Docking brings in a GPU client then I will no longer devote my computing power to this project.
Why??
My CPU cycles, my choice...
That's an incredibly unctuous position to take, assuming you know all the facts.
Your credit is rudimentially based off of how much 'work' is performed. It has NOTHING to do with the 'type' of processor you use to complete it. Did you know that most projects would not benefit any more with a GPU app, then they would a CPU app? - Just because there are projects out there that are lucky enough to have the type of work that flourishes on a GPU, doesn't mean that every project will bear the same fruit.
The only logical 'Non-inflammatory' version of deciphering why you would want to leave this project would be that you don't have a GPU to devote to docking@home, are assuming that a Docking GPU application would swallow your position in the leaderboard with GPU-enabled machines,
and
you are
graciously
going to offer your CPU to a project that lacks a GPU application.
When presented *as i now have articulated it* this is a truly altruistic and wonderful thing for a person to do - HOWEVER Your candor and 'tune' tell us, this is certainly
not
the case.
I'm going to just blankly *assume* that since you listed your current rank, along with your statement that you are selfishly devoting your resources to this project solely so you can climb a stats chart(?!), and are now THREATENING the project with your 'defection' should we open up a possibly stronger avenue of research. ---- Well... now that we know that you could give a crap about the HIV(and other) science being done here: I'll freely concede that stats are important to everyone, However -*THIS ARGUMENT ASIDE*- If another avenue of gaining credit presents itself, what's so hard about opening up your computer to a little GPU work? - Technology changes. Change with it, or get left behind. <-- Oh, and this
isn't
your choice.
I hear this same crap everytime a new version of windows comes out, or a new car hits the market. I suppose when we all get complacent with what we got we can have our whoopie goldberg moment, but your troll-worthy comment is ultimately thoughtless, just like her.
Oh and by the way - Your paltry 4 million credits will pale in comparison to what my GPU array will do, were we to lose you. I personally don't care about the credits, but
god forbid we actually focus on the science being performed and curing diseases rather than our own egos.
Now I don't want to scare you away from projects, but what I would
strongly advise you to do
is to find a project you feel passionately about. As you can tell - This one is mine. I could care less if I am in
last
place, I am here because I believe in what this project is trying to do. I hope that one day, something that we do here (possibly through even my own machine) will help in the fight against HIV. Since you clearly are more interested in your credits rather than innovating, adapting, or merely contributing what you can - I really think you need to find projects you actually care about. - That way when things like this change that push you down the stats list, you can still feel good about yourself without stroking the ego button.
Sorry, but if Docking brings in a GPU client then I will no longer devote my computing power to this project.
Why??
My CPU cycles, my choice...
That's an incredibly unctuous position to take, assuming you know all the facts.
Your credit is rudimentially based off of how much 'work' is performed. It has NOTHING to do with the 'type' of processor you use to complete it. Did you know that most projects would not benefit any more with a GPU app, then they would a CPU app? - Just because there are projects out there that are lucky enough to have the type of work that flourishes on a GPU, doesn't mean that every project will bear the same fruit.
The only logical 'Non-inflammatory' version of deciphering why you would want to leave this project would be that you don't have a GPU to devote to docking@home, are assuming that a Docking GPU application would swallow your position in the leaderboard with GPU-enabled machines,
and
you are
graciously
going to offer your CPU to a project that lacks a GPU application.
When presented *as i now have articulated it* this is a truly altruistic and wonderful thing for a person to do - HOWEVER Your candor and 'tune' tell us, this is certainly
not
the case.
I'm going to just blankly *assume* that since you listed your current rank, along with your statement that you are selfishly devoting your resources to this project solely so you can climb a stats chart(?!), and are now THREATENING the project with your 'defection' should we open up a possibly stronger avenue of research. ---- Well... now that we know that you could give a crap about the HIV(and other) science being done here: I'll freely concede that stats are important to everyone, However -*THIS ARGUMENT ASIDE*- If another avenue of gaining credit presents itself, what's so hard about opening up your computer to a little GPU work? - Technology changes. Change with it, or get left behind. <-- Oh, and this
isn't
your choice.
I hear this same crap everytime a new version of windows comes out, or a new car hits the market. I suppose when we all get complacent with what we got we can have our whoopie goldberg moment, but your troll-worthy comment is ultimately thoughtless, just like her.
Oh and by the way - Your paltry 4 million credits will pale in comparison to what my GPU array will do, were we to lose you. I personally don't care about the credits, but
god forbid we actually focus on the science being performed and curing diseases rather than our own egos.
Now I don't want to scare you away from projects, but what I would
strongly advise you to do
is to find a project you feel passionately about. As you can tell - This one is mine. I could care less if I am in
last
place, I am here because I believe in what this project is trying to do. I hope that one day, something that we do here (possibly through even my own machine) will help in the fight against HIV. Since you clearly are more interested in your credits rather than innovating, adapting, or merely contributing what you can - I really think you need to find projects you actually care about. - That way when things like this change that push you down the stats list, you can still feel good about yourself without stroking the ego button.
Lot of assumptions in your post..... Credit isn't everything
My participation in Docking is nothing at all to do with ego. My wife lost a good friend to AIDS.
Some time ago, one of the student in the group looked at how GPUs could help volunteer computing and tried to make some predictions. We wrote a short paper that you can download and read at:
It's a VERY good idea to work on gpu code, using the vast computational power of modern gpu card.
But i prefer if you use opencl (i know, cuda is more mature) because: a) it's open and b) is compatible with all hw (the other side of the moon, amd).
Have you idea about gpu version? Roadmap?
ID:
6311 | Rating: 0
| rate:
/
Michela
Forum moderator
Project administrator
Project developer
Project tester
Project scientist
It's a VERY good idea to work on gpu code, using the vast computational power of modern gpu card.
But i prefer if you use opencl (i know, cuda is more mature) because: a) it's open and b) is compatible with all hw (the other side of the moon, amd).
Have you idea about gpu version? Roadmap?
We are working with CUDA because when we started the coding, it was more mature.
You are right, OpenCL code is portable across GPUs and CPUs. At the same time OpenCL is not as fast as CUDA at this point. The OpenCL community is making interesting progress and we are observing and considering the next move.
____________
If you are interested in working on Docking@Home in a great group at UDel, contact me at 'taufer at acm dot org'!
We are working with CUDA because when we started the coding, it was more mature.
You are right, OpenCL code is portable across GPUs and CPUs. At the same time OpenCL is not as fast as CUDA at this point. The OpenCL community is making interesting progress and we are observing and considering the next move.
I assume that your are using BOINC because you have a enormous amount of work to be done. And so your desire for speed is not to finish a job as fast as possible but to finish as much work as possible in a given amount of time. Given these assumptions it is hard to believe that you are maximizing work done by focusing on speed using CUDA and excluding so many GPUs by not using OpenCL.
Docking@home display Wu and use GPU ? like Folding@home ?
@+
*_*
No, we are not distributing jobs on GPUs at this point. One of the reasons is that CHARMM does not run on GPUs. We look at GPUs with interest and have been working on a MD code for GPUs. One of the students in the group is testing a small framework with BOINC and our MD code for GPUs. This could become a very interesting follow-up for Docking@Home but we will need some time.
____________
If you are interested in working on Docking@Home in a great group at UDel, contact me at 'taufer at acm dot org'!
We look at GPUs with interest and have been working on a MD code for GPUs. One of the students in the group is testing a small framework with BOINC and our MD code for GPUs. This could become a very interesting follow-up for Docking@Home but we will need some time.
I've found the the versions of BOINC available so far do not offer much support for OpenCL GPU workunits. There are plans to add such support in the 6.14.* series of versions, though. The versions so far seem to offer adequate support for OpenCL CPU workunits, though.
Milkyway@Home is using their own ways of filling in the missing pieces of OpenCL support; looks adequate if you don't have multiple GPUs on the same computer that you want to share among different BOINC projects offering GPU workunits, but likely to need many changes when the 6.14.* versions of BOINC become available.
I've found a program described as offering some help for translating CUDA to OpenCL, but not able to do the full job.