Appears that they don't agree with what I saw previously about what I saw about a 6.14.* series. Not very clear if they have simply renumbered that series 7.0.*, though.
Here's another question about repeated messages from the project.
"7/24/2011 12:06:59 PM | Einstein@Home | Your app_info.xml file doesn't have a version of Gravitational Wave S6 GC search ."
I always get this message whenever I contact the project. The question I have is why I get the message at all since I only crunch GPU work and have the project preferences stating only GPU work and no work or applications for CPU.
When I look at my project preferences page now, I see that it says that I requested to process Gravitational Wave S6 GC work. I never set that preference. Also, when I try to edit my preferences for this project, I am unable to..... the option is always greyed out with no ability to change the preference to no. How does this get fixed?
When I look at my project preferences page now, I see that it says that I requested to process Gravitational Wave S6 GC work. I never set that preference. Also, when I try to edit my preferences for this project, I am unable to..... the option is always greyed out with no ability to change the preference to no. How does this get fixed?
Keith
This will not be fixed, because the main purpose of E@H is the search for gravitational waves. This was mentioned from Bruce Allen again, just a couple of weeks ago. I only can't find the thread at the moment to provide a link.
Well that response really sucks. So I have to live with the annoying messages until further notice? Why don't the project programmers come up with a CUDA app for gravitational wave searches. I'd be happy to run it on my GPU's. It isn't going to happen on any CPU's.
Do the preferences at least allow a selection for CPU work only or GPU work only? If they do, wouldn't it be adequate for the project to supply a suitable app_info.xml section for CPU Gravitational Wave S6 GC, have you add it, and then never send you workunits for it?
Well that response really sucks. So I have to live with the annoying messages until further notice? Why don't the project programmers come up with a CUDA app for gravitational wave searches. I'd be happy to run it on my GPU's. It isn't going to happen on any CPU's.
Keith
GPUs are only effective with apps that either have tiny memory footprints or extremely sequential memory access patterns; the GW app's access patterns are too random to work on a GPU. Back when nVidia first launched CUDA, they attempted to port several large distributed computing projects to use it. E@H was one of them, but they were unable to get it running faster than a CPU app, and the memory trashing meant that it was just as slow on a top end GPU as a mid level card. Short of coming up with a completely new way to analyze the data a GPU app won't be happening anytime soon.
I've seen something on another BOINC project saying that for the AMD/ATI Radeon series of graphics boards, they found that the HD 5000 series of boards offered significantly faster access to larger groups of memory than the HD 4000 series and lower.
Does that make it look like a good time for another evaluation for the HD 5000 series boards, and any other AMD/ATI boards with GPUs similar enough? If the source code and input files for the previous attempts were saved, just recompiling it to run only on the higher level boards should be sufficient for enough of a new evaluation to see if any more effort is worthwhile.
In case the earlier source code is for CUDA only, I've found some references to a Swan program for converting CUDA to OpenCL.
The current versions of BOINC make it difficult to offer OpenCL GPU workunits, but there are plans to change that in a a series of versions after 6.12.*. Still, would an evaluation that did not use BOINC be adequate for determining if any more effort in this direction is now worthwhile?
If the application uses much double precision, the Milkyway@Home project has already shown that it's worth separate evaluations for the high-end AMD/ATI GPUs that support double precision, and the more common ones that don't.
Also, was the Nvidia GPU evaluation done before the Fermi chips became available? Most, possibly all, of the Fermi chips support double precision.
Do the preferences at least allow a selection for CPU work only or GPU work only? If they do, wouldn't it be adequate for the project to supply a suitable app_info.xml section for CPU Gravitational Wave S6 GC, have you add it, and then never send you workunits for it?
Yes, the E@H project preferences page has a setting for GPU work only and CPU work only. I have GPU work only set for Yes and CPU work set for NO. I don't think the scheduler has that capability of not sending out work units for a valid application in the project folder.
Quote:
Einstein@Home preferences
Resource share
Determines the proportion of your computer's resources allocated to this project. Example: if you participate in two BOINC projects with resource shares of 100 and 200, the first will get 1/3 of your resources and the second will get 2/3. 100
Use CPU
Enforced by version 6.10+ no
Use NVIDIA GPU
Enforced by version 6.10+ yes
Is it OK for Einstein@Home and your team (if any) to email you? yes
Should Einstein@Home show your computers on its web site? yes
Default computer location home
Graphics setting: frames per second (FPS)
Warning: affects CPU consumption! Default value: 20 20
Graphics setting: render quality
Warning: requires hardware 3D acceleration! Default value: low low
Graphics setting: window width (pixels)
Default value: 800 800
Graphics setting: window height (pixels)
Default value: 600 600
Run only the selected applications Binary Radio Pulsar Search: yes
Binary Radio Pulsar Search: yes
Gravitational Wave S6 GC search : yes
Gamma-ray pulsar search #1: no
Run CPU versions of applications for which GPU versions are available no
Do the preferences at least allow a selection for CPU work only or GPU work only? If they do, wouldn't it be adequate for the project to supply a suitable app_info.xml section for CPU Gravitational Wave S6 GC, have you add it, and then never send you workunits for it?
Yes, the E@H project preferences page has a setting for GPU work only and CPU work only. I have GPU work only set for Yes and CPU work set for NO. I don't think the scheduler has that capability of not sending out work units for a valid application in the project folder.
Quote:
Einstein@Home preferences
Resource share
Determines the proportion of your computer's resources allocated to this project. Example: if you participate in two BOINC projects with resource shares of 100 and 200, the first will get 1/3 of your resources and the second will get 2/3. 100
[snip]
Run CPU versions of applications for which GPU versions are available no
Edit Einstein@Home preferences
I've found that once you complete enough GPU workunits successfully, BOINC will notice that it needs to give preference to BOINC projects offering only CPU workunits whenever it is ready to ask for more CPU workunits. Therefore, giving enough of a resource share to BOINC projects that offer only CPU workunits, and have a reliable supply of those workunits, should combine with the above settings so that BOINC stops asking Einstein@Home for any more CPU workunits, even if Einstein@Home would still send some CPU workunits if it was ever asked for any.
Appears that they don't agree
)
Appears that they don't agree with what I saw previously about what I saw about a 6.14.* series. Not very clear if they have simply renumbered that series 7.0.*, though.
Here's another question about
)
Here's another question about repeated messages from the project.
"7/24/2011 12:06:59 PM | Einstein@Home | Your app_info.xml file doesn't have a version of Gravitational Wave S6 GC search ."
I always get this message whenever I contact the project. The question I have is why I get the message at all since I only crunch GPU work and have the project preferences stating only GPU work and no work or applications for CPU.
I only crunch BRP3/4 CUDA work.
Cheers, Keith
When I look at my project
)
When I look at my project preferences page now, I see that it says that I requested to process Gravitational Wave S6 GC work. I never set that preference. Also, when I try to edit my preferences for this project, I am unable to..... the option is always greyed out with no ability to change the preference to no. How does this get fixed?
Keith
RE: When I look at my
)
This will not be fixed, because the main purpose of E@H is the search for gravitational waves. This was mentioned from Bruce Allen again, just a couple of weeks ago. I only can't find the thread at the moment to provide a link.
Well that response really
)
Well that response really sucks. So I have to live with the annoying messages until further notice? Why don't the project programmers come up with a CUDA app for gravitational wave searches. I'd be happy to run it on my GPU's. It isn't going to happen on any CPU's.
Keith
Do the preferences at least
)
Do the preferences at least allow a selection for CPU work only or GPU work only? If they do, wouldn't it be adequate for the project to supply a suitable app_info.xml section for CPU Gravitational Wave S6 GC, have you add it, and then never send you workunits for it?
RE: Well that response
)
GPUs are only effective with apps that either have tiny memory footprints or extremely sequential memory access patterns; the GW app's access patterns are too random to work on a GPU. Back when nVidia first launched CUDA, they attempted to port several large distributed computing projects to use it. E@H was one of them, but they were unable to get it running faster than a CPU app, and the memory trashing meant that it was just as slow on a top end GPU as a mid level card. Short of coming up with a completely new way to analyze the data a GPU app won't be happening anytime soon.
I've seen something on
)
I've seen something on another BOINC project saying that for the AMD/ATI Radeon series of graphics boards, they found that the HD 5000 series of boards offered significantly faster access to larger groups of memory than the HD 4000 series and lower.
Does that make it look like a good time for another evaluation for the HD 5000 series boards, and any other AMD/ATI boards with GPUs similar enough? If the source code and input files for the previous attempts were saved, just recompiling it to run only on the higher level boards should be sufficient for enough of a new evaluation to see if any more effort is worthwhile.
In case the earlier source code is for CUDA only, I've found some references to a Swan program for converting CUDA to OpenCL.
http://www.cse.scitech.ac.uk/disco/mew20/presentations/GPU_MattHarvey.pdf
http://www.khronos.org/news/permalink/swan-command-line-tool-port-cuda-to-opencl/
The current versions of BOINC make it difficult to offer OpenCL GPU workunits, but there are plans to change that in a a series of versions after 6.12.*. Still, would an evaluation that did not use BOINC be adequate for determining if any more effort in this direction is now worthwhile?
If the application uses much double precision, the Milkyway@Home project has already shown that it's worth separate evaluations for the high-end AMD/ATI GPUs that support double precision, and the more common ones that don't.
Also, was the Nvidia GPU evaluation done before the Fermi chips became available? Most, possibly all, of the Fermi chips support double precision.
RE: Do the preferences at
)
Yes, the E@H project preferences page has a setting for GPU work only and CPU work only. I have GPU work only set for Yes and CPU work set for NO. I don't think the scheduler has that capability of not sending out work units for a valid application in the project folder.
RE: RE: Do the
)
I've found that once you complete enough GPU workunits successfully, BOINC will notice that it needs to give preference to BOINC projects offering only CPU workunits whenever it is ready to ask for more CPU workunits. Therefore, giving enough of a resource share to BOINC projects that offer only CPU workunits, and have a reliable supply of those workunits, should combine with the above settings so that BOINC stops asking Einstein@Home for any more CPU workunits, even if Einstein@Home would still send some CPU workunits if it was ever asked for any.