Nbody Release 1.54
log in

Advanced search

Message boards : News : Nbody Release 1.54

Author Message
Sidd
Project developer
Project tester
Project scientist
Send message
Joined: 19 May 14
Posts: 53
Credit: 101,723
RAC: 0

Message 64087 - Posted: 11 Nov 2015, 17:45:10 UTC

Hey All,

I just released a new version of nbody, version 1.54. There are no major changes, just some minor bug fixes. There was a sampling bug in the velocity assignment which caused it to throw away more possible vels than it should. So the initialization is much faster than before. However, the results should be the same. But faster is always better!

Let me know if there is any issues with the new version/binaries.

Thank you!
Sidd

Zaacharia
Send message
Joined: 15 Sep 15
Posts: 1
Credit: 37,117
RAC: 24

Message 64090 - Posted: 12 Nov 2015, 20:19:01 UTC

I don't know if it is related but I am not getting any work now - don't know when it started but I left with running climateprediction until I get more work.

Profile Wisesooth
Send message
Joined: 2 Oct 14
Posts: 26
Credit: 9,552,659
RAC: 7,169

Message 64102 - Posted: 16 Nov 2015, 15:46:15 UTC

I do not know if this has anything to do with nBody, but Cortana and Start menu have failed twice recently in Windows 10 while running BOINC. System logged me off and back in again when I pressed Microsoft's "fixit" button. BOINC did not like that at all. Milkyway tasks restarted from last checkpoint. Looks to me like Cortana is misbehaving when BOINC is running. I have no Cortana mischief in other Windows 10 systems. Thought you might like to know.

BTW Sidd, I started to work on a cool GO button for you. See Heidi for details.
____________

Profile Wisesooth
Send message
Joined: 2 Oct 14
Posts: 26
Credit: 9,552,659
RAC: 7,169

Message 64104 - Posted: 16 Nov 2015, 16:11:15 UTC - in response to Message 64090.

Sounds familiar. I fat fingered my mouse when it happened to hover over "Send no more work" button when I first started. Check if you have something wrong in your global preferences.

If you are on the BOINC Basic window, click on ADVANCED. Next, click on the PROJECTS button at the top. It will show a list of projects you chose to run. Click on the Milkyway project to highlight it. A pane will show on your left. Go to the Milkyway home page (the bottom button). Once you are there, scroll down until you see PROFILES and launch that. Look on the bottom of the next screen and select "My Account." (NOTE: I did not design this.)

You will see two preference buttons. You want to open "Preferences for this project." The next screen shows a list of task types you gave permissions to download. They have check boxes on the left of each task type. If none of them are checked, that is your problem. Look for a tiny "edit" button and select it. Now you can check the boxes for the task types you want to run. Click on SAVE and close the page. That should get you back to the PROJECT window. Click on UPDATE (the top button). When it finishes doing its thing, click on the TASKS pane. Hopefully, you will see Milkyway tasks flowing into your list.

Hope this helps you, and thank you for crunching for milkyway.
A fellow cruncher,
____________

sjmielh
Send message
Joined: 9 Aug 09
Posts: 4
Credit: 2,086,496
RAC: 701

Message 64130 - Posted: 29 Nov 2015, 10:36:47 UTC - in response to Message 64087.

Thanks for this update Sidd

Now my Macs can process NBody again! I still get some errors, but it doesn't run past 100%

Sjmielh

Profile [AF>Le_Pommier] Jerome_C2005
Send message
Joined: 1 Apr 08
Posts: 19
Credit: 388,333
RAC: 101

Message 64165 - Posted: 11 Dec 2015, 21:27:52 UTC
Last modified: 11 Dec 2015, 21:43:36 UTC

Please read in the other topic for announcement of n-body 1.52 what I wrote (one month ago and now again), nobody ever answered me but they are not working on my Mac : running on 1 core out of 8, blocking the other boinc apps to run (pretending it is "mt"), and in 1.52 they would not even complete normally (AND they started running again now when my setup does not allow nbody)...

http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=3819&postid=64164


EDIT : there is a change now, after 15mn of running they actually start to run on all 8 cores and really using 100% of my CPU (instead of 1/8th).


EDIT2 : they don't completely run 100%, it's more like 90% of available CPU, letting a 10% idle.


EDIT3 : OK so the task did finish on estimated time and was sent successfully back to the servers, so things look better with this 1.54, even though not perfect (running 15mn with 1/8 of CPU and not running a full 100% for the remaining time).

Captiosus
Send message
Joined: 9 Apr 14
Posts: 17
Credit: 797,083
RAC: 14

Message 64178 - Posted: 15 Dec 2015, 16:28:02 UTC - in response to Message 64165.

Please read in the other topic for announcement of n-body 1.52 what I wrote (one month ago and now again), nobody ever answered me but they are not working on my Mac : running on 1 core out of 8, blocking the other boinc apps to run (pretending it is "mt"), and in 1.52 they would not even complete normally (AND they started running again now when my setup does not allow nbody)...

http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=3819&postid=64164


EDIT : there is a change now, after 15mn of running they actually start to run on all 8 cores and really using 100% of my CPU (instead of 1/8th).


EDIT2 : they don't completely run 100%, it's more like 90% of available CPU, letting a 10% idle.


EDIT3 : OK so the task did finish on estimated time and was sent successfully back to the servers, so things look better with this 1.54, even though not perfect (running 15mn with 1/8 of CPU and not running a full 100% for the remaining time).

It because, as Sidd said, the initialization period cannot be made multi-threaded. It screws up the math needed to set the initial values of each body in the model, and when it gets turned in along with other workunits from the same batch, the results poorly correlate due to the bad math.

My suggestion in response to that was to use the idle initialization period to prime a batch of workunits for compute, then once enough are primed and ready to go, process them one by one.
Alternatively, interleave it by setting 87.5% of the available threads for MT (on my CPU, it'd be 14 threads of 16 total, on yours it'd be 7 of 8 total), and use the remaining thread(s) to initialize the next unit, so when an MT workunit is done, another one can immediately start processing.

Leo
Send message
Joined: 17 May 09
Posts: 2
Credit: 10,767,457
RAC: 2,710

Message 64189 - Posted: 18 Dec 2015, 17:42:38 UTC - in response to Message 64087.

I am running Windows 10 and all of a sudden my tasks say GPU missing and will abort. Graphics card is a GTX 780.

Any idea?

Thanks

Richard Haselgrove
Send message
Joined: 4 Sep 12
Posts: 218
Credit: 448,778
RAC: 0

Message 64190 - Posted: 18 Dec 2015, 18:22:42 UTC - in response to Message 64189.

N-Body tasks won't be even trying to use your GPU.

Try reading here.

Profile [AF>Le_Pommier] Jerome_C2005
Send message
Joined: 1 Apr 08
Posts: 19
Credit: 388,333
RAC: 101

Message 64192 - Posted: 18 Dec 2015, 20:48:23 UTC

Thanks for the explanation captiosus.

Leo
Send message
Joined: 17 May 09
Posts: 2
Credit: 10,767,457
RAC: 2,710

Message 64198 - Posted: 20 Dec 2015, 16:23:07 UTC - in response to Message 64190.

Thanks. Everything is working now. One thing I had to do was download a program, Driver Fusion, to remove all of the Nvidia drivers including the ones that came with Windows 10. Installed the latest Nvidia drivers and all is working fine.

Thanks again.

marmot
Send message
Joined: 12 Dec 15
Posts: 2
Credit: 10,050
RAC: 0

Message 64251 - Posted: 16 Jan 2016, 15:23:22 UTC
Last modified: 16 Jan 2016, 15:29:16 UTC

I have a 6 core n-body WU that has run for 3 days and 10 hours and is only at 48% completion. The other 6 core n-body's in que report estimates of 38 to 58 minutes. Is a 7 day run time possible? Will this 7 day WU get paid an appropriate amount of credit? Should I abort this WU?

I can't make a determination of normality because your server deletes my results so quickly. It would be much appreciated if you would maintain 2 weeks of results in our account history so we can get an idea of when packets failed and which machines are under-performing or if a WU app is behaving badly.

Besides this extrremely long calculation time, I've noticed with n-body that sometimes the 8 core n-body will be running and along with another 2 WU's from other projects even though BOINC only thinks this machine has 8 cores. It seems the n-body WU doesn't suspend when BOINC does a WU switch over every 30 minutes.

My configuration is probably rare. Many of my machines are set to a cc_config.xml options <ncpus>N</ncpus> where N is 2 cores higher than actual system cores. It's the only solution that actually fixes the work fetch anomaly where BOINC debt/workfetch algorithm idles a core (or 2) so that a high resource project with no current WU's has a core ready to go. I see this work fetch problem on many of my machines that have Citizen Grid (or a few other intermittent projects) set to 99 resource while the 6 or 8 other projects are set to 20 or less resource share. All the real cores are kept working 24/7 and when a intermittent high priority project actually gets work fetched the BOINC virtual cores get that WU and the OS deals with the extra thread sharing.

Is the <ncpus>N</ncpus> > than real ncpu's an issue for the n-body app?

hristo
Send message
Joined: 19 Nov 14
Posts: 3
Credit: 59,059
RAC: 48

Message 64299 - Posted: 4 Feb 2016, 21:25:35 UTC

Hello Sidd,

my comment is written here http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=3885&postid=64297

about some handles leak in windows xp 64 bit.


Post to thread

Message boards : News : Nbody Release 1.54


Main page · Your account · Message boards


Copyright © 2016 AstroInformatics Group