Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
![]() |
World Community Grid Forums
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 16
|
![]() |
Author |
|
grumpydaddy
Cruncher Joined: Aug 6, 2012 Post Count: 5 Status: Offline |
I suspect this is an issue with current HCC beta tests however ....
----------------------------------------Pic posted on XS forum: This sure looks like a GFAM WU is trying to use GPU ![]() Thoughts anyone? EDIT: Link: http://imageshack.us/photo/my-images/827/gfamgpu.jpg/ [Edit 2 times, last edit by grumpydaddy at Sep 7, 2012 5:23:04 PM] |
||
|
ryan222h
Senior Cruncher Joined: Sep 4, 2006 Post Count: 425 Status: Offline |
Many of the projects currently being run on wcg would not be suitable for a GPU application, not to mention it takes a huge amount of programming and planning to make it possible.
----------------------------------------I have confidence that researchers are looking at all the best options for solving their problems and they will choose the right path for their project. ![]() |
||
|
raybon
Cruncher Joined: Dec 2, 2005 Post Count: 29 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Definitely looks like it. I've been seeing BETA_HCC units recently, but the unit in the picture isn't marked BETA. I'm guessing something wonky may be going on with his BOINC client. If he hadn't trimmed so much off the bottom you would see his client version. I don't think there are any plans to develop a GPU application for GFAM with such a short expected run time.
----------------------------------------![]() |
||
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3295 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Something similar hapenned with CFSW. It's not a GPU app for GFAM.
----------------------------------------AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
Rickjb
Veteran Cruncher Australia Joined: Sep 17, 2006 Post Count: 666 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
The only current WCG project which could benefit from GPU processing is HCC, according to a post some time ago by SekeRob.
HCC involves image processing, and that's of course what GPUs were designed for. The only GPU work so far has been several rounds of HCCGPU beta tests - see the beta test sections of the WCG forum. They are having lots of problems working with the variety of GPU hardware out there. Somewhere I read that the programmers of the Rosetta program, which is used at WCG for HPF2, tried a novel approach to port it it to GPUs. However, this was abandoned because they found that there would have been no speed improvement even if they had managed to iron out all the bugs. My suggestion for WCG speedups would be to make maximum use of the mutimedia/vector/SIMD instructions (SSE, AVX etc) built into X86 CPUs. Every modern x86 CPU has at least some of these instructions, while only a small proportion of computers on the grid have or ever will have suitable GPUs and users willing to use them for WCG. Perhaps some WCG projects already use these extended instructions, but no-one at WCG answers my enquiries about this. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello Rickjb,
Fortunately, any 64-bit compiler can assume all the SSE instructions and use libraries specialized for them. This would speed things up if the original project software were programmed to take advantage of this. My understanding (which might well be wrong) is that very little of our project software was written this way. The BLAS libraries allow programs in many languages to be optimized for streaming extensions almost as though they were written in APL. Which would organize the program in a way that could be ?easily? adapted to GPU coding (or quickly show why that algorithm should not be used on GPUs). As far as I know, none of our project software uses the Basic Linear Algebra Subroutine libraries. It is disappointing but WCG generally gets involved long after scientists have roughed out a program design and decided it is good enough to try to obtain massive computer resources to run it. Also, as a bald ancient programmer, I suspect that today's kids do not think of efficiency when they start planning their design. ![]() Lawrence |
||
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
My suggestion for WCG speedups would be to make maximum use of the mutimedia/vector/SIMD instructions (SSE, AVX etc) built into X86 CPUs. While a discussion about using SSE and so on can be highly interesting, I fail to see how your post is even remotely connected to this thread... Remember, the OP basically said: "Why do GFAM suddenly use a GPU to crunch?" Now, clearly, this is either due to a client-bug or a server-bug. On the client-side, as is often the case the OP failed to specify the client-version. Still, it's more likely it's a server-side bug. During the beta there's been a random chance of getting a 293-error, and this seems to have been tracked to a scheduler-bug. By having an error that ocassionally forgets something in scheduler-reply, it's also possible the same bug ocassionally includes too much in other scheduler-replies, including demanding GFAM to use a GPU. So, hopefully the problem is now fixed, and grumpydaddy and other users won't hit the same problem with GFAM and other non-GPU-projects again. ![]() "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
Norton01
Cruncher US Joined: Feb 19, 2012 Post Count: 41 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Hi 1st post here
----------------------------------------I just wanted to confirm that this GFAM work unit is running on my rig using the GPU (0.40 CPU + 1.00 ATI GPU's) GFAM_x1Q4Jtetramer_PfGST_0034803_0139 They have been running for a couple of days now and there are more waiting to run. They are actually running on an extra thread too (FX-8150 with 9 wu's running rather than 8) Rig specs are an FX-8150 w/ AMD HD 7850 GPU, etc...
TPU-WCG Team Captain
WCG at TechPowerUp! http://www.techpowerup.com/forums/forums/world-community-grid-wcg.68/ |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello Norton01,
No WCG project uses a GPU. GOFAM certainly does not. WCG occasionally runs a Beta test on a variant of HCC that does use a GPU. It has not been released as a normal project yet. Lawrence |
||
|
Norton01
Cruncher US Joined: Feb 19, 2012 Post Count: 41 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Thank you for your reply :)
----------------------------------------I'm running a 9th thread on my 8 core cpu and this 9th thread is always stated the same way/same wu. See pic: ![]() I admit I don't know that much about how these projects use PC resources but I'm producing valid results using this 9th thread and have no other way to explain it???
TPU-WCG Team Captain
----------------------------------------WCG at TechPowerUp! http://www.techpowerup.com/forums/forums/world-community-grid-wcg.68/ [Edit 1 times, last edit by Norton01 at Sep 8, 2012 11:44:12 PM] |
||
|
|
![]() |