| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Beta Testing
Forum: Beta Test Support Forum
Thread: New Beta Test - February 25, 2016 [ Issues Thread ] |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 91
|
|
| Author |
|
|
foxfire
Advanced Cruncher United States Joined: Sep 1, 2007 Post Count: 121 Status: Offline Project Badges: 
|
I need this result from a wu that has not been suspended The last 3 betas I finished were not restarted and ended normally. They went to 100%, paused a few seconds, uploaded and reported complete. BETA_HST1_000002_000874_AC0010_F00074_S00001_0 BETA_HST1_000002_000299_AC0004_F00099_S00001_0 BETA_HST1_000002_000560_AC0007_F00060_S00001_0 I have 3 more that I will try restarting to see how they end. It will take a while to report as I have a number of CEP betas to complete on the same device.  |
||
|
|
Seoulpowergrid
Veteran Cruncher Joined: Apr 12, 2013 Post Count: 815 Status: Offline Project Badges: 
|
Got 4 WUs on 2.66 GHz Ubuntu boxes. All 4 are in PV, 3 took 21 hours and 4th took 22 hours.
----------------------------------------Wingmen took 13 hours to 23 hours to complete on same WUs.  |
||
|
|
foxfire
Advanced Cruncher United States Joined: Sep 1, 2007 Post Count: 121 Status: Offline Project Badges:
|
After reading OldChaps post I restarted 2 of the HST wus and observed:
----------------------------------------The wu restarts from the next to last checkpoint The time remaining and percent complete are garbage following the restart The wu appears to run correctly following the restart, but there is no way to monitor its progress unless you look at the stderr.txt file in the slots folder. ------------------------------------------------------------- Properties following restart: CPU time last checkpoint: 01d,02:33:25 CPU time: 01d,02:326:27 Elapsed time: 01d,02:48:29 Estimated Time remaining: 00:00:00 Fraction done: 842.626% ------------------------------------------------------------- stdout.txt after suspend-before restart No previous checkpoint file present, assuming this is a new run. [21:08:12] [INFO] Checkpoint complete. .... [21:38:07] [INFO] Checkpoint complete. [21:58:36] [INFO] Checkpoint complete. ------------------------------------------------------------- stderr.txt file after suspend-before restart INFO: No state to restore. Start from the beginning. [20:57:59] INFO: Running initial simulatoin Writting checkpoint at step 590. [21:15:15] INFO: Completed step 1000 of initial simulation Writting checkpoint at step 1170. ... [21:32:55] INFO: Completed step 89000 of initial simulation Writting checkpoint at step 89460. ------------------------------------------------------------- stderr.txt after restart [21:53:19] INFO: Running initial simulatoin Reading checkpoint file state.cpt generated: Sat Feb 27 21:38:07 2016 Writting checkpoint at step 89820. [22:01:35] INFO: Completed step 90000 of initial simulation Writting checkpoint at step 90420. [22:18:13] INFO: Completed step 91000 of initial simulation Writting checkpoint at step 91020. .... [23:41:04] INFO: Completed step 96000 of initial simulation Writting checkpoint at step 96450. ------------------------------------------------------------- |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7545 Status: Offline Project Badges: 
|
Funny how they misspell "writing" as "writting." I am assuming the examples were copied and pasted directly from BOINC, not written personally by Foxfire.
----------------------------------------Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
That'd be likely from the HST app as others are not 'writting', but writing :o)
----------------------------------------BTW, the author may have been thinking of ´writhing´ as needing contortions to generate a checkpoint ;O))))) Think it's time to start setting up HST1 in the Charting DB... inserting a data block I do so rarely, it's going to be rust all over to be removed first.  [Edit 1 times, last edit by SekeRob* at Feb 28, 2016 2:01:30 PM] |
||
|
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: 
|
----------------------------------------
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
  |
||
|
|
foxfire
Advanced Cruncher United States Joined: Sep 1, 2007 Post Count: 121 Status: Offline Project Badges:
|
Funny how they misspell "writing" as "writting." I am assuming the examples were copied and pasted directly from BOINC, not written personally by Foxfire. Cheers Correct, cut & paste from the .txt files |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Interesting article about GROMACS and GPU acceleration. http://www.gromacs.org/GPU_acceleration Interesting article indeed, but I'm not sure how it would be used with BOINC. Aren't we forced into one-task-per-CPU-thread operation, rather than have a single task use all the CPU threads available? And how would that work if there was a GPU-enabled version? It would seem in that case to also want to use all the CPU threads as well -- or did I miss something? BTW, I got CEP2 betas but none of these so I've got nothing to look at. It might just be chance, or maybe my old machine is too slow. I didn't notice any minimum requirements being published, but maybe I missed that too! (Perhaps I need new glasses, though a new brain would probably be more beneficial.) |
||
|
|
SekeRob
Master Cruncher Joined: Jan 7, 2013 Post Count: 2741 Status: Offline |
Relies on the implementation... for the former HCC on GPGPU, you could allocate a fraction of one CPU core/thread per HCC job on GPGPU i.e depending on what card and the rest, you could run many on one host and still have CPU threads free to run CPU only jobs. There's quite a good control to be achieved combining settings in cc_config and app_config.
|
||
|
|
Mumak
Senior Cruncher Joined: Dec 7, 2012 Post Count: 477 Status: Offline Project Badges:
|
BOINC supports multi-threaded tasks (MT) as well.
----------------------------------------But I see no reason that such application would require MT, generally any GPU task requires a certain amount of CPU resources to 'feed' the GPU. Some applications (like the HCC1 in the past) might also require a certain pre-/post-processing done by the CPU (a part of code that's not possible to port into GPGPU). Anyway, I think it's too early for such speculations. WCG won't certainly comment anything in advance ;-) We can only hope that some time we will get a notice of a new GPU Beta test ;-) Edit: Gromacs supports OpenCL too and it seems to offload only certain parts of code to GPUs, the rest is still done on CPU.  [Edit 1 times, last edit by Mumak at Feb 28, 2016 8:09:41 PM] |
||
|
|
|