Binary Radio Pulsar Search (Parkes PMPS XT) "BRP6"

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2143
Credit: 2960792669
RAC: 706403

Sorry, v1.49 crashes on my

Sorry, v1.49 crashes on my test machine without the --device safety net, runs OK with it.

Task 487600769 has the same basic error as Gavin, with debug output - that's a failure from initial start. I have two others which crashed on resumption from checkpoint when I restarted without --device.

Going off to see what happens on machines which ran v1.47 without error - I suspect the difference is that they only have a single GPU.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5872
Credit: 117769392014
RAC: 34782197

Looks like 1.49 might have

Looks like 1.49 might have already been withdrawn. I've noticed a couple of hosts that were getting 1.49 a while ago have just now started getting 1.39 again :-(.

Cheers,
Gary.

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250661459
RAC: 34531

Yep, still a bug in there.

Yep, still a bug in there. Building new one.

BM

BM

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2143
Credit: 2960792669
RAC: 706403

Seeing v1.39 again here

Seeing v1.39 again here too.

v1.49 runs OK under WinXP/32 and older BOINC v6.12.34 (hardly surprising)

v1.49 fails under Win7/32 and BOINC v7.4.36, single GPU - same error, task 487602339. That host ran v1.47 as originally deployed.

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2143
Credit: 2960792669
RAC: 706403

BTW, information for other

BTW, information for other testers: if you have been adding a --device safety net via app_config.xml, it will get added to new app versions as released. You have to remove it from client_state manually for a true test: simply removing it from app_config isn't enough. (Though possibly adding a null command line to app_config might do it - I'll test that next time)

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5872
Credit: 117769392014
RAC: 34782197

I reported this information

I reported this information earlier in this thread.

Quote:


[pre]
Search App Ver Elapsed time CPU time Notes
====== ======= ============ ========= =====
BRP5 1.39 28,500 sec 3,750 sec Long term averages for BRP5
BRP6 1.41 38,500 sec 4,950 sec Averages for 20+ tasks for BRP6 v1.41
BRP6-beta 1.47 21,980 sec 2,230 sec 1st task - 18% on v1.41 and 82% on v1.47 app
BRP6-beta 1.47 20,567 sec 2,027 sec 2nd task - 100% on v1.47 app
BRP6-beta 1.47 19,053 sec 1,760 sec 3rd task - 100% on v1.47 app
BRP6-beta 1.47 22,509 sec 2,405 sec 4th task - 100% on v1.47 app

BRP6-beta 1.47 19,099 sec 2,072 sec 1st v1.47 task on different host but identical hardware

BRP6-beta 1.47 17,880 sec 664 sec Pentium dual core G3258 (Haswell refresh) with HD7850 4x[/pre]

The nice thing is that these old clunkers are now able to get performance on the HD7850 that is much closer to what a Haswell refresh based system can do. I just grabbed a recently completed result from such a machine and put it in the above table for comparison.


There has now been sufficient time for the machine doing the above 4 1.47 tasks to have done a total of 31 such tasks. I found an online statistics calculator and fed the 31 elapsed times into it and got the following information back.

Mean = 21317 secs
Std Dev = 5135 secs
Variance = 26377229

For the statistically challenged (like me :-).) it might be easier to report that the bulk of the results were clustered around 19,000 to 20,000 secs with only 7 of the 31 that exceeded 25,000 secs. There were also 8 results below 19,000 secs. The actual lowest value was 18,680 and the highest was 32,242.

Considering that this host used to average around 28,500 for BRP5, I'm mightily pleased with how it's going on BRP6-Beta. My heartfelt thanks to all those involved in the development and deployment of this brilliant new app.

Cheers,
Gary.

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250661459
RAC: 34531

RE: New (BRP6 Beta) CUDA

Quote:
New (BRP6 Beta) CUDA version 1.49 is out. This is meant to fix the API problem, no change to the actual computation code.

Another go: 1.50 is out (CUDA only, computation code is still identical to 1.47).

BM

BM

Gavin
Gavin
Joined: 21 Sep 10
Posts: 191
Credit: 40644337738
RAC: 1

Thanks Bernd, I have one

Thanks Bernd,

I have one running on my previously mentioned host with a GTX660Ti, 10 minutes in and looking very good against the two v1.39's that are also running.

Gavin.

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2143
Credit: 2960792669
RAC: 706403

Also running fine here on my

Also running fine here on my GTX 670 testbed - task 487745609 - with no command line.

Well done to Bernd and the team.

Bent Vangli
Bent Vangli
Joined: 6 Apr 11
Posts: 23
Credit: 725742660
RAC: 0

Running one 1.50 at a GTX 680

Running one 1.50 at a GTX 680 and one at a GTX 980 Linux 64 bit (Ubuntu). Both seems fine so far. They are running faster than the 1.39. Good work Bernd. :-)

Bent

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.