Docking at home and CUDA GPU ?
Message boards : Number crunching : Docking at home and CUDA GPU ?
| Author | Message | |
|---|---|---|
|
http://setiweb.ssl.berkeley.edu/beta/cuda.php
|
||
ID:
4648 | Rating: 0
| rate:
/
|
||
|
I'll be happy to contribute my GPU to this project. I'm hoping there will be a Docking application for Linux 32-bit architecture. |
||
|
ID:
4902 | Rating: 0
| rate:
/
|
||
|
GPUs are rather difficult to rewrite the programs for; so far, I've only found two BOINC projects which have done it. The result, though, is much faster computing.
|
||
|
ID:
4934 | Rating: 0
| rate:
/
|
||
|
Worth giving ATIGPU a look?
|
||
|
ID:
5373 | Rating: 0
| rate:
/
|
||
|
Hi, quite new to DOCKING, started with SETI, GPUgrid; EINSTEIN; Leiden Classical and LHC, which had no work since april or so.
|
||
|
ID:
5407 | Rating: 0
| rate:
/
|
||
|
Hi, quite new to DOCKING, started with SETI, GPUgrid; EINSTEIN; Leiden Classical and LHC, which had no work since april or so. Sorry Fred, but AQUA does not use ATI (CAL). It did have an CUDA application but has now stopped running it as the use of multiple CPUs were more efficient and ran the jobs faster. Milkyway uses ATI (CAL) and CUDA, Collatz uses ATI (CAL) and CUDA as well. GPUGRID, SETI, Einstein all have CUDA applications and there are a few more projects that are getting applications rewritten to also use CUDA and maybe ATI (CAL). The credit awarded was increased to make it more inline with other projects as it has been a consistent low payer for awhile, it coincided with a new fixed credit system. Conan. ____________ 
|
||
|
ID:
5413 | Rating: 0
| rate:
/
|
||
|
The last time I checked, both yoyo@home and The Lattice Project had essentially stopped their efforts to use GPUs.
|
||
|
ID:
5421 | Rating: 0
| rate:
/
|
||
|
Towards Large-Scale Molecular Dynamics
|
||
|
ID:
5445 | Rating: 0
| rate:
/
|
||
|
Towards Large-Scale Molecular Dynamics This best I can tell, they have a paper on their plans to offer GPU workunits in the future, with no signs that they have actually started offering any of them now. p.s.: As Yoyo is a transport vehicle for 4 other (non-BOINC) projects, their GPU support depends on the GPU support of the science projects. Distributed.net does have a CUDA client but afaik. that's for rc5, not for ogr I already have an account at Yoyo as well, and have seen no signs that they even list rc5 among the 4 sub-projects they are currently offering workunits for, at least on the English version of their web pages. My guess is that Yoyo's wrapper program isn't able to handle GPU applications yet. p.p.s.: drugdiscoveryathome has a CUDA client for one of their applications I happen to have an account there as well (just started in the last few days), and if I've gotten any GPU workunits from them yet, they made no effort to make it clear that they were GPU workunits. Also, they still call themselves a pre-alpha BOINC project, and the rather high error rate for their workunits should convince most people to agree with that. For example, a large percentage of them include input files that don't even download correctly. I've already tried the The Lattice Project project. Even though they have an application named GARLI-GPU, it appears to actually be a CPU-only type of workunit intended to gather some information about which of the participating computers can support GPU use, and any plans to turn it into an application that actually uses GPUs to do anything useful appear to have been abandoned there, at least until Nvidia and/or ATI offer better software support, such as a compiler for the computer language that project is already using. I'll pass them the information on Nvidia's plans to offer GT300 chips and some new compilers soon, and ask them to check if the new compilers will work for the Nvidia boards already available. The Rosetta at home and Superlink@Technion projects also appear to have abandoned any efforts to offer a GPU version for now; I'll make sure both know about the new compilers as well. |
||
|
ID:
5452 | Rating: 0
| rate:
/
|
||
|
Yes, the group is working in porting MD to GPU but it will take some time before we get to a point of doing docking on GPU. It's in our plans but for a long term period of time.
|
||
|
ID:
5489 | Rating: 0
| rate:
/
|
||
|
Hi all!
|
||
|
ID:
5596 | Rating: 0
| rate:
/
|
||
|
Have you checked if Harvard offers any online courses that would give me sufficient access to the source code for the CHARMM program you use to have a try at using my years of Fortran experience to make the changes needed to allow it to be compiled using a Fortran to CUDA compiler, in order for you to be ready to offer suitable workunits at least for Nvidia GPUs?
|
||
|
ID:
5646 | Rating: 0
| rate:
/
|
||
|
I did a Google-equivalent search for Fortran to OpenCL compilers, and saw no sign that anybody is even mentioning one in progress on their web sites yet.
|
||
|
ID:
5668 | Rating: 0
| rate:
/
|
||
|
GPUGRID is now trying to get started using AMD GPUs. However, there's a good chance that they will not be able to use any older than the very new HD 7xxx series (first released earlier this year). It seems that it's the first series AMD has released that included some hardware features they need.
|
||
|
ID:
6558 | Rating: 0
| rate:
/
|
||
|
I hope I will be able to use this: "CAL AMD Radeon HD 6x00 series (Turks) (1024MB) driver: 1.4.1646"
|
||
|
ID:
6559 | Rating: 0
| rate:
/
|
||
|
Are the plans to have a GPU application someday still on? Wish there was OpenCL app eventually. |
||
|
ID:
6619 | Rating: 0
| rate:
/
|
||
|
Not in our lifetime. :P |
||
|
ID:
6689 | Rating: 0
| rate:
/
|
||
|
ATI support wouldnt be to bad too :P |
||
|
ID:
6690 | Rating: 0
| rate:
/
|
||
|
I hope I will be able to use CAL AMD Radeon HD 6x00 series (Turks) (1024MB) |
||
|
ID:
6720 | Rating: 0
| rate:
/
|
||
|
Yes, the group is working in porting MD to GPU but it will take some time before we get to a point of doing docking on GPU. It's in our plans but for a long term period of time. Post from admin - 28 Oct 2009!!! I think 918 days is enough to tell us something new..... |
||
|
ID:
6721 | Rating: 0
| rate:
/
|
||
|
For those that keep asking about doing Docking using GPU there is an alternative project that is working similar science - GPUGRID.
|
||
|
ID:
6723 | Rating: 0
| rate:
/
|
||
|
Hi guys,
|
||
|
ID:
6734 | Rating: 0
| rate:
/
|
||
|
Hmm what we need is a programmer guru to spend a month or two working on it with the docking guys and gals. Anyone know anyone who fancies doing something good for science, for free? |
||
|
ID:
6735 | Rating: 0
| rate:
/
|
||
|
Hmm what we need is a programmer guru to spend a month or two working on it with the docking guys and gals. Anyone know anyone who fancies doing something good for science, for free? I might, AFTER I get information on which computer languages are involved and make sure that I have studied those computer languages. It will have to be telecommuting rather than in person, though. |
||
|
ID:
6736 | Rating: 0
| rate:
/
|
||
|
Hmm what we need is a programmer guru to spend a month or two working on it with the docking guys and gals. Anyone know anyone who fancies doing something good for science, for free? Both CUDA (nvidia) and OpenCL (ATI/nvidia/CPU) programming is typically done in C/C++, although there are interfaces to allow programming in other languages. But C/C++ is the main language. Despite using a familiar language, there's a bit of a learning curve involved for both architectures. If you want to go all-out nuts for the fastest execution it's possible to program the CUDA (nvidia) cards in their native assembly language. I have no idea what the Docking application is written in. |
||
|
ID:
6737 | Rating: 0
| rate:
/
|
||
|
(duplicate) |
||
|
ID:
6738 | Rating: 0
| rate:
/
|
||
|
Hmm what we need is a programmer guru to spend a month or two working on it with the docking guys and gals. Anyone know anyone who fancies doing something good for science, for free? I've taken a C++ course lately, am now studying C from a book, and am planning to go on to CUDA and OpenCL later, so I think that I'm approaching being ready. I've done some assembly language programming, but at least 20 years ago; I'm in no hurry to try that again. I have three computers available, all running 64-bit Windows. No experience in Linux at all. I do have Cygwin installed, which tries to emulate Linux under Windows. Where can I download the source code for the Docking application so I can check for anything else I need to study before starting? Are there any license restrictions on who can see the source code? |
||
|
ID:
6739 | Rating: 0
| rate:
/
|
||
|
I have 2 ATI 4670 1 GB DDR3-1333 in CrossFire (only 32-bit floating point), I have joined but I'll contribute my GPUs only, my CPU is for WCG. SO if you ever have a stable, efficient and optimized GPU for CAL/OpenCL 1.0b (beta - 32-bit floats) , which is the best my ATI cards can do, I'll be happy to help you.
|
||
|
ID:
6745 | Rating: 0
| rate:
/
|
||
|
Hi guys, Are you interested in comunity help? Can you publish your source code? |
||
|
ID:
6746 | Rating: 0
| rate:
/
|
||
|
Are you interested in comunity help? Not interested? |
||
|
ID:
6775 | Rating: 0
| rate:
/
|
||
|
Are you interested in offering an online course in the appropriate GPU programming language so that an online student can try to produce a GPU version? |
||
|
ID:
6776 | Rating: 0
| rate:
/
|
||
|
Are you interested in offering an online course in the appropriate GPU programming language so that an online student can try to produce a GPU version? There are many people that know how to write gpu code (student, pro, etc), ATI and Nvidia help developers and there is a lot of resouces on the net (http://www.multicoreinfo.com/2009/08/parprog-part-9/ for example), but without source code , it's impossible to create any gpu app |
||
|
ID:
6777 | Rating: 0
| rate:
/
|
||
|
Hi guys, May 2012 News??? |
||
|
ID:
7131 | Rating: 0
| rate:
/
|
||
|
I've now finished one online CUDA course and started another. I'm looking for an online OpenCL course, but haven't found one yet.
|
||
|
ID:
7132 | Rating: 0
| rate:
/
|
||
|
A GPU version of CHARMM already exists, but does not appear to be in use at Docking@Home.
|
||
|
ID:
7137 | Rating: 0
| rate:
/
|
||
|
Could we have an explanation of why it's not used yet? Because Docking are using a 5-years-old version of Charmm.... |
||
|
ID:
7146 | Rating: 0
| rate:
/
|
||
|
The team continue to publish paper about gpu
|
||
|
ID:
7212 | Rating: 0
| rate:
/
|
||
|
The team continue to publish paper about gpu I suppose it's all because this project's team is way too small, henceforth it wouldn't be possible for them to process such amount of results - basically humans are the bottleneck here :P Also remember that this server sometimes barely can handle the CPU work units throughput. Poof, another bottleneck. I suppose this project just has a really low priority @ udel. Feels more like some side-project that allows them to show-off like "look, we're doing these things, yeah 'important' research and distributed computing!", and I'm not bashing the project team here, rather their superiors/public organisations funding them. |
||
|
ID:
7214 | Rating: 0
| rate:
/
|
||
Message boards : Number crunching : Docking at home and CUDA GPU ?
Database Error: The MySQL server is running with the --read-only option so it cannot execute this statement
array(3) {
[0]=>
array(7) {
["file"]=>
string(47) "/boinc/projects/docking/html_v2/inc/db_conn.inc"
["line"]=>
int(97)
["function"]=>
string(8) "do_query"
["class"]=>
string(6) "DbConn"
["object"]=>
object(DbConn)#43 (2) {
["db_conn"]=>
resource(168) of type (mysql link persistent)
["db_name"]=>
string(7) "docking"
}
["type"]=>
string(2) "->"
["args"]=>
array(1) {
[0]=>
&string(51) "update DBNAME.thread set views=views+1 where id=373"
}
}
[1]=>
array(7) {
["file"]=>
string(48) "/boinc/projects/docking/html_v2/inc/forum_db.inc"
["line"]=>
int(60)
["function"]=>
string(6) "update"
["class"]=>
string(6) "DbConn"
["object"]=>
object(DbConn)#43 (2) {
["db_conn"]=>
resource(168) of type (mysql link persistent)
["db_name"]=>
string(7) "docking"
}
["type"]=>
string(2) "->"
["args"]=>
array(3) {
[0]=>
object(BoincThread)#3 (16) {
["id"]=>
string(3) "373"
["forum"]=>
string(1) "2"
["owner"]=>
string(4) "2291"
["status"]=>
string(1) "0"
["title"]=>
string(30) "Docking at home and CUDA GPU ?"
["timestamp"]=>
string(10) "1391211680"
["views"]=>
string(4) "2007"
["replies"]=>
string(2) "37"
["activity"]=>
string(33) "0.0000000000000021037342878677003"
["sufferers"]=>
string(1) "0"
["score"]=>
string(1) "0"
["votes"]=>
string(1) "0"
["create_time"]=>
string(10) "1229212416"
["hidden"]=>
string(1) "0"
["sticky"]=>
string(1) "0"
["locked"]=>
string(1) "0"
}
[1]=>
&string(6) "thread"
[2]=>
&string(13) "views=views+1"
}
}
[2]=>
array(7) {
["file"]=>
string(63) "/boinc/projects/docking/html_v2/user/community/forum/thread.php"
["line"]=>
int(184)
["function"]=>
string(6) "update"
["class"]=>
string(11) "BoincThread"
["object"]=>
object(BoincThread)#3 (16) {
["id"]=>
string(3) "373"
["forum"]=>
string(1) "2"
["owner"]=>
string(4) "2291"
["status"]=>
string(1) "0"
["title"]=>
string(30) "Docking at home and CUDA GPU ?"
["timestamp"]=>
string(10) "1391211680"
["views"]=>
string(4) "2007"
["replies"]=>
string(2) "37"
["activity"]=>
string(33) "0.0000000000000021037342878677003"
["sufferers"]=>
string(1) "0"
["score"]=>
string(1) "0"
["votes"]=>
string(1) "0"
["create_time"]=>
string(10) "1229212416"
["hidden"]=>
string(1) "0"
["sticky"]=>
string(1) "0"
["locked"]=>
string(1) "0"
}
["type"]=>
string(2) "->"
["args"]=>
array(1) {
[0]=>
&string(13) "views=views+1"
}
}
}
query: update docking.thread set views=views+1 where id=373