Message boards :
News :
nbody simulation assimilator up and running
Message board moderation
Author | Message |
---|---|
Send message Joined: 30 Aug 07 Posts: 2046 Credit: 26,480 RAC: 0 |
It looks like the nbody simulation assimilator is up and going. I know the workunits are really short, but right now we're just testing to see if the searches can get to the known best position and trying to debug the assimilator. It's a lot easier when we have a bit more of a work flow going. Once we have things a bit more debugged, the nbody simulation workunit times should increase quite dramatically. Right now we're simulating around 4000 bodies, but we expect to increase that to around 100,000 to 1,000,000 bodies (or more). |
Send message Joined: 20 Mar 09 Posts: 7 Credit: 1,016,945 RAC: 0 |
I just processed over 20 WUs that all "crashed on take-off", i.e. returned a computing error almost immediately after starting. The error files seem to all contain: "dyld: unknown required load command 0x80000022" For example, de_nbody_test_10_477_1281947127_0. NorPer It looks like the nbody simulation assimilator is up and going. I know the workunits are really short, but right now we're just testing to see if the searches can get to the known best position and trying to debug the assimilator. It's a lot easier when we have a bit more of a work flow going. |
Send message Joined: 8 May 10 Posts: 576 Credit: 15,979,383 RAC: 0 |
Yes, that problem has hypothetically been solved. Are you on and old version of OS X? It seems to be because by default, statically linked binaries on 10.6 don't work on older versions of OS X. |
Send message Joined: 20 Mar 09 Posts: 7 Credit: 1,016,945 RAC: 0 |
I'm running 10.5.8. Will the new WUs eventually work with an OS older than 10.6 or should I take my upgrade DVD closer to the top of my pile of "things to do" on my desk? |
Send message Joined: 8 May 10 Posts: 576 Credit: 15,979,383 RAC: 0 |
It should work eventually. I added something which might have fixed it last night, but I don't know if Travis has updated the binaries being sent out yet. I also don't have an old OS X to test on, so I don't know if it actually works yet. However, when the OpenCL applications are ready at some point in the future, you'll need 10.6 to use those. |
Send message Joined: 14 Feb 09 Posts: 999 Credit: 74,932,619 RAC: 0 |
Just crashed the app on my system, or exceeded the max time. http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=176156983 DCF is at 1.00 as I reset the project just before I got this unit. It was estimated at 9:08 when downloaded but crashed at 18:30. |
Send message Joined: 30 Apr 09 Posts: 101 Credit: 29,871,946 RAC: 334 |
http://milkyway.cs.rpi.edu/milkyway/apps.php The new MilkyWay@Home N-Body Simulation apps will replace the old MilkyWay@Home apps ? Or later they will run simultaneously? For CPU (& simultaneously GPU) cruncher.. The current available opt. CPU apps: http://www.arkayn.us/milkyway/index.html ..are still well? Later also? Will we see later also opt. N-Body apps for CPU? Or they will be the fastest available (no space for optimization), like the stock GPU apps currently are? Thanks! |
Send message Joined: 30 Aug 07 Posts: 2046 Credit: 26,480 RAC: 0 |
http://milkyway.cs.rpi.edu/milkyway/apps.php They'll be going simultaneously. There's different research being done with the applications.
Hopefully we'll be able to get the nbody applications nice and optimized ourselves so other people don't have to do it. Also, the code for milkyway@home is now all open source on github: http://github.com/Milkyway-at-home/milkywayathome_client So if any of you guys come up with code optimizations it'd be nice if you'd share them with us so everyone can benefit. We also can add people who are interested in working on the code to the project so they can submit improvements and optimizations and maybe even get their name of the milkyway@home people page as independent contributors. :P --Travis |
Send message Joined: 18 Nov 07 Posts: 280 Credit: 2,442,757 RAC: 0 |
I know this is still in the testing phase, but could you add an option to the Milkyway@Home Preferences (the "Preferences for this project") to only crunch for the nbody search? I wouldn't be adding much trying to compete with the GPUs on the normal WUs. |
Send message Joined: 30 Aug 07 Posts: 2046 Credit: 26,480 RAC: 0 |
I know this is still in the testing phase, but could you add an option to the Milkyway@Home Preferences (the "Preferences for this project") to only crunch for the nbody search? I wouldn't be adding much trying to compete with the GPUs on the normal WUs. I'll look into it, not quite sure how to do it to be honest. Although it doesn't seem like it would be too difficult. |
Send message Joined: 18 Nov 07 Posts: 280 Credit: 2,442,757 RAC: 0 |
I'll look into it, not quite sure how to do it to be honest. Although it doesn't seem like it would be too difficult. Thanks :) If it turns out to be difficult to figure out, the Einstein@Home scientists should be able to help out; they allow you to disable crunching for the "Arecibo Binary Pulsar Search (STSP)" (though I wouldn't want to in that case, especially given their recent discovery!) |
Send message Joined: 19 Feb 08 Posts: 350 Credit: 141,284,369 RAC: 0 |
I know this is still in the testing phase, but could you add an option to the Milkyway@Home Preferences (the "Preferences for this project") to only crunch for the nbody search? I wouldn't be adding much trying to compete with the GPUs on the normal WUs. Hi, one idea to make this to an auto-setup: If a GPU-version for an app is available and the pc reports a valid gpu, cpu-apps of this type are automatic disabled. If no gpu-app is available, cpu-apps are are enabled by default. This could make things easier, especially if new apps arise or currently inactive apps (mw3) become active again. Regards, Alexander |
Send message Joined: 30 Aug 07 Posts: 2046 Credit: 26,480 RAC: 0 |
I know this is still in the testing phase, but could you add an option to the Milkyway@Home Preferences (the "Preferences for this project") to only crunch for the nbody search? I wouldn't be adding much trying to compete with the GPUs on the normal WUs. Well, I posted on the BOINC mailing lists, and DA said it would be non-trivial to implement. But I think we'll have GPU versions of the nbody simulation application out pretty soon anyways, so hopefully it won't be that big of a deal. |
Send message Joined: 19 Feb 08 Posts: 350 Credit: 141,284,369 RAC: 0 |
Hi Emanuel, maybe they could, but do they want to help? I checked the gpu-load with the cuda app of e@h, it uses ~9% of the gpu. So I posted a request for help for an app_info, which allows to set the nr. of gpu's to 0.5 to run 2 apps at a time. And of course, it should include the cpu-apps they want us to run. This was three weeks ago, no answer. I posted a request for help for a mw app_info yesterday, and its running since 10 hours! What a difference! I posted the request also in the seti-germany forum, one reply supporting my request but nothing else. If one of the xml-gurus here also crunches e@h, maybe he could help? Kind regards, Alexander |
Send message Joined: 5 Apr 09 Posts: 71 Credit: 6,120,786 RAC: 0 |
Hello you can help me with the file to receive app_info uts 0.21 and n-body, thank you !? <app_info> <app> <name>milkyway</name> </app> <file_info> <name>astronomy_0.21_x64_SSE3.exe</name> <executable/> </file_info> <app_version> <app_name>milkyway</app_name> <version_num>21</version_num> <file_ref> <file_name>astronomy_0.21_x64_SSE3.exe</file_name> <main_program/> </file_ref> </app_version> <app_version> <app_name>milkyway</app_name> <version_num>20</version_num> <file_ref> <file_name>astronomy_0.21_x64_SSE3.exe</file_name> <main_program/> </file_ref> </app_version> <app> <name>milkyway</name> </app> <file_info> <name>astronomy_0.21_x64_SSE3.exe</name> <executable/> </file_info> Team Alliance francophone, boinc: 7.0.18 GA-P55-UD5, i7 860, Win 7 64 bits, 8g DDR3, GTX 470 |
Send message Joined: 30 Apr 09 Posts: 101 Credit: 29,871,946 RAC: 334 |
Why you didn't asked me? ;-) AFAIK, it should be very easy to do this.. At SETI@home you go to the entry in your app_info.xml file: <coproc> <type>CUDA</type> <count>1</count> </coproc> and edit it to: <coproc> <type>CUDA</type> <count>0.5</count> </coproc> And you would see 2 WUs/GPU. It should be the same at Einstein@home. I don't crunch Einstein@home, so I can't give you a complete app_info.xml file. But, AFAIK (not for sure), you would need at least WinVista for to see 2+ WUs/GPU. |
Send message Joined: 30 Apr 09 Posts: 101 Credit: 29,871,946 RAC: 334 |
I could post my MilkyWay@home app_info.xml . I don't know how are your experiences about stock and optimized project apps. You could take the MW@h app_info.xml and edit it for Einstein@home. For example, MW@h have two, SETI@home have three: <file_ref> <file_name>xxxxxx.dll</file_name> </file_ref> ..entries for the CUDA app. You need to edit all in red. You need maybe edit all in blue. <app_info> <app> <name>milkyway</name> <user_friendly_name>MilkyWay@home optimized</user_friendly_name> </app> <file_info> <name>milkyway_optimized_0.21_SSE3.exe</name> <executable/> </file_info> <app_version> <app_name>milkyway</app_name> <version_num>19</version_num> <file_ref> <file_name>milkyway_optimized_0.21_SSE3.exe</file_name> <main_program/> </file_ref> </app_version> <app> <name>milkyway</name> </app> <file_info> <name>milkyway_0.24_windows_intelx86__cuda23.exe</name> <executable/> </file_info> <file_info> <name>cudart.dll</name> <executable/> </file_info> <file_info> <name>cutil32.dll</name> <executable/> </file_info> <app_version> <app_name>milkyway</app_name> <version_num>24</version_num> <plan_class>cuda23</plan_class> <avg_ncpus>0.050000</avg_ncpus> <coproc> <type>CUDA</type> <count>1</count> </coproc> <file_ref> <file_name>milkyway_0.24_windows_intelx86__cuda23.exe</file_name> <main_program/> </file_ref> <file_ref> <file_name>cudart.dll</file_name> </file_ref> <file_ref> <file_name>cutil32.dll</file_name> </file_ref> </app_version> </app_info> For not to hijack more this thread, we could continue via PM. Because it's E@h topic. The best would be over the S@h site. There I'm every day. :-) |
Send message Joined: 17 Oct 08 Posts: 36 Credit: 411,744 RAC: 0 |
Is there any problem with the validation currently? It appears all results are going into pending, even those with quorum=1 and not counting those with errors of course, and the status remains 'Completed, waiting for validation'. Regards |
Send message Joined: 19 Feb 08 Posts: 350 Credit: 141,284,369 RAC: 0 |
Looks like your'e right. The nbody's are all waiting even if the wingman has completed his wu. The error-rate is extreme high, more than 30% at the moment. Maybe they are trying some reworks now. Alexander |
Send message Joined: 19 Feb 08 Posts: 350 Credit: 141,284,369 RAC: 0 |
This is why I love this forum! CU there, Alexander |
©2024 Astroinformatics Group