APv6.01 via Lunatics 0.40 installer has errored out all APv6 tasks

Message boards : Number crunching : APv6.01 via Lunatics 0.40 installer has errored out all APv6 tasks
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · Next

AuthorMessage
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1211354 - Posted: 29 Mar 2012, 1:10:24 UTC

Anyone know what is going on with these so far returned AP 6.01 tasks?

http://setiathome.berkeley.edu/result.php?resultid=2371339987
http://setiathome.berkeley.edu/result.php?resultid=2371081282

I used the Lunatics 0.40 installer and everything seemed to go all right with the other apps. Is there some part of the installer that didn't get put in for the AstroPulse 6.01 tasks?

Keith
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1211354 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1211358 - Posted: 29 Mar 2012, 1:27:25 UTC - in response to Message 1211354.  

Anyone know what is going on with these so far returned AP 6.01 tasks?

http://setiathome.berkeley.edu/result.php?resultid=2371339987
http://setiathome.berkeley.edu/result.php?resultid=2371081282

I used the Lunatics 0.40 installer and everything seemed to go all right with the other apps. Is there some part of the installer that didn't get put in for the AstroPulse 6.01 tasks?

Keith

Raistmer's r555 running on your AMD Phenom(tm) II X6 1100T, and failing after 4-6 hours? I don't think we saw anything like that in testing. Interesting (and different) error messages in both cases - I'll ask him to take a look.

You'll probably want to suspend the other 8 AP_v6 tasks you have in progress while he investigates (provided you have enough other work to occupy your CPUs - otherwise, let them run, as the additional output could help diagnosis).
ID: 1211358 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1211364 - Posted: 29 Mar 2012, 1:43:41 UTC - in response to Message 1211358.  

Thanks for the quick response Richard. I think I will suspend the AP 6.01 tasks until I here different. I have plenty of other work. The beta NV AP 5.05 app is still returning good results. On that note ...... is this app ever going to make it out of beta?

Keith
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1211364 · Report as offensive
Cosmic_Ocean
Avatar

Send message
Joined: 23 Dec 00
Posts: 3027
Credit: 13,516,867
RAC: 13
United States
Message 1211394 - Posted: 29 Mar 2012, 3:40:55 UTC

Something else that may possibly be useful for diagnosis is to make a copy of those WUs and see if they do the same thing in stand-alone mode outside of BOINC. Just a suggestion.
Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)
ID: 1211394 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1211396 - Posted: 29 Mar 2012, 3:41:52 UTC

Richard, something else must be going on with the installer. I just updated my other computer with the Lunatics 0.40 installer and now that computer is erroring out the CPU tasks.

http://setiathome.berkeley.edu/result.php?resultid=2373727865
http://setiathome.berkeley.edu/result.php?resultid=2373727863
http://setiathome.berkeley.edu/result.php?resultid=2373727861
http://setiathome.berkeley.edu/result.php?resultid=2373727859
http://setiathome.berkeley.edu/result.php?resultid=2373727858
http://setiathome.berkeley.edu/result.php?resultid=2373727856
http://setiathome.berkeley.edu/result.php?resultid=2373727854

Can you backlevel to the old 0.39 installer?

Keith

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1211396 · Report as offensive
JLConawayII

Send message
Joined: 2 Apr 02
Posts: 188
Credit: 2,840,460
RAC: 0
United States
Message 1211405 - Posted: 29 Mar 2012, 4:05:02 UTC

Congratulations on finding the only broken .40 installer in the known universe. You win the prize! Your super-secret surprise will arrive in the mail in a few days.




(it's not really a secret, it's a dead skunk I found along the highway)
ID: 1211405 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1211406 - Posted: 29 Mar 2012, 4:05:21 UTC - in response to Message 1211404.  

Richard, my mistake using the installer. I thought I had properly toggled the radio button for the AMD app. But after comparing the correctly running app on the first computer, noticed the other one had installed the Intel app. I just redid the installer and checked the app_info and see that it has the correct AMD app now. Sorry, for the panic.

Keith

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1211406 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1211409 - Posted: 29 Mar 2012, 4:09:16 UTC - in response to Message 1211394.  

Not sure how to do that exactly. Guess you turn off network communications and run through the task without letting it report?

Keith

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1211409 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1211488 - Posted: 29 Mar 2012, 10:35:00 UTC - in response to Message 1211414.  

Should have been more explicit. Meant how to run the AP6.01 tasks offline. I don't need to backlevel the Lunatics app since the errored tasks were caused by a mistake in setup. Just have to deal with the problem of errored tasks and resends always caused by my inevitable typo in modifying app_info and then dumping my cache.

Another question. Are the AP6.01 tasks very different from the AP5.05 ones? Are they doing a lot more science? I was wondering why their completion times are so much longer than AP5.05 tasks. Looks like it will take quite a bit of time to build up 10 completed tasks in order to generate a APR that I can then use to input a valid FLOPS the application.

Keith

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1211488 · Report as offensive
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 34253
Credit: 79,922,639
RAC: 80
Germany
Message 1211529 - Posted: 29 Mar 2012, 14:32:39 UTC
Last modified: 29 Mar 2012, 15:06:56 UTC

No AP 6.01 is not much different.
The calcultion is done different and outlier detection added but both apps r555 and r557 are definetly faster than r409.
On my machine ~35% - 45%.


With each crime and every kindness we birth our future.
ID: 1211529 · Report as offensive
LadyL
Volunteer tester
Avatar

Send message
Joined: 14 Sep 11
Posts: 1679
Credit: 5,230,097
RAC: 0
Message 1211571 - Posted: 29 Mar 2012, 15:59:38 UTC

Keeping tasks for offline runs: look into you project dir and copy the tasks to someplace else. If we have the task links, we can grab the task from the server, but with the line that saturated...

Offline runs of suspect tasks are best done by somebody else and not on the affected host. Not that I think something is wrong wioth the task itself.

Completion times should actually be faster than r409. Another indication something is wrong.

For the umptiest time: the NVidia OpenCL app will make it out of beta and into the installer, when it doesn't hog a CPU core on last generation drivers.
I'm not the Pope. I don't speak Ex Cathedra!
ID: 1211571 · Report as offensive
Profile skildude
Avatar

Send message
Joined: 4 Oct 00
Posts: 9541
Credit: 50,759,529
RAC: 60
Yemen
Message 1211574 - Posted: 29 Mar 2012, 16:00:20 UTC
Last modified: 29 Mar 2012, 16:03:33 UTC

Avx takes 15-16 hours on my FX-8150. Thats significantly faster than the SSE or is it SSE2.

It appears to be about 20-40K faster than the stock app on an i7 860



http://setiathome.berkeley.edu/workunit.php?wuid=953354362


In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope
ID: 1211574 · Report as offensive
Cosmic_Ocean
Avatar

Send message
Joined: 23 Dec 00
Posts: 3027
Credit: 13,516,867
RAC: 13
United States
Message 1211576 - Posted: 29 Mar 2012, 16:10:13 UTC - in response to Message 1211488.  

Should have been more explicit. Meant how to run the AP6.01 tasks offline.

PMed.
Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)
ID: 1211576 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1211737 - Posted: 30 Mar 2012, 2:06:00 UTC - in response to Message 1211576.  

I noticed that there a couple of new files in the Seti directory after the installer finished. First there is something called a "wisdom" file. Second, there is a dll called libfftw3f-3 that I don't remember being there from the 0.39 installer. I wonder if the issue with the failed AP6.01 tasks was that this file wasn't loaded. If I remember generally, you don't load a .DLL file until you reboot Windows, which I did not do, I just stopped BOINC, ran the installer and restarted BOINC. I didn't see any mention of rebooting Windows after the install in the release notes. Maybe I was supposed to reboot as a general principle.

I decided to restart the suspended AP6.01 tasks to see if they would properly complete. I seem to have successfully finished one task with a couple of other failed. The failed ones might have been worked on before I rebooted Windows and maybe the .DLL issue was to blame.

I don't see any improvement in completion times with AP6.01 app compared to the old 5.05 app. I completed the successful task in about 50K seconds, right about where AP5.05 tasks finished with the Lunatics CPU app from the 0.39 installer.

I will continue to crunch AP6.01 tasks unless I start seeing every task failing.

Cheers, Keith
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1211737 · Report as offensive
Cosmic_Ocean
Avatar

Send message
Joined: 23 Dec 00
Posts: 3027
Credit: 13,516,867
RAC: 13
United States
Message 1211775 - Posted: 30 Mar 2012, 4:48:44 UTC

The libfftw DLL is static/built-in to the optimized AP apps, but I guess it is still included/there as a back-up?
Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)
ID: 1211775 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1211792 - Posted: 30 Mar 2012, 5:51:42 UTC - in response to Message 1211775.  

The libfftw DLL is static/built-in to the optimized AP apps, but I guess it is still included/there as a back-up?

r555 has dynamic linkage vs FFTW. So FFTW DLL isrequired for it to operate properly. For GPU builds too.

Linkage was changed in my builds to reflect multicore nature of today PCs.
Shared via DLL code can reside in common cache. Also, it will use less memory. Will these effects affect performance and in what degree - not known currently.
ID: 1211792 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1211851 - Posted: 30 Mar 2012, 8:57:12 UTC - in response to Message 1211737.  

Second, there is a dll called libfftw3f-3 that I don't remember being there from the 0.39 installer. I wonder if the issue with the failed AP6.01 tasks was that this file wasn't loaded. If I remember generally, you don't load a .DLL file until you reboot Windows...

No, that only applies if you're updating previous DLLs which are being used by an application or process which can't be be stopped - and that usually means only core Windows components. Because we can (and do) stop both BOINC and the science applications when the installer runs, we'll never need you to reboot Windows just for Lunatics.

You may be confusing program installation with Windows Update, which does need to work on core components.
ID: 1211851 · Report as offensive
LadyL
Volunteer tester
Avatar

Send message
Joined: 14 Sep 11
Posts: 1679
Credit: 5,230,097
RAC: 0
Message 1211862 - Posted: 30 Mar 2012, 9:40:02 UTC
Last modified: 30 Mar 2012, 9:40:48 UTC

The wisdom file is generated by the app - something to do with picking the best fft out of the lib. Up to now it was generated in the slot directory.
Not having to generate it because you can get one from the projet dir saves some time.
I'm not the Pope. I don't speak Ex Cathedra!
ID: 1211862 · Report as offensive
tbret
Volunteer tester
Avatar

Send message
Joined: 28 May 99
Posts: 3380
Credit: 296,162,071
RAC: 40
United States
Message 1211869 - Posted: 30 Mar 2012, 10:53:22 UTC - in response to Message 1211574.  
Last modified: 30 Mar 2012, 11:15:08 UTC

AMD 8120. - First AP 6, 10% blanked, 46k sec.

I have no idea if that's "good" or even "better" but I thought I'd share in case anyone's curious about AVX.


Edit: as opposed to an 1100T, 17.5% blanked at 51.1k.
ID: 1211869 · Report as offensive
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 34253
Credit: 79,922,639
RAC: 80
Germany
Message 1211906 - Posted: 30 Mar 2012, 14:29:00 UTC - in response to Message 1211869.  
Last modified: 30 Mar 2012, 14:29:21 UTC

AMD 8120. - First AP 6, 10% blanked, 46k sec.

I have no idea if that's "good" or even "better" but I thought I'd share in case anyone's curious about AVX.


Edit: as opposed to an 1100T, 17.5% blanked at 51.1k.


Thats not bad tbret.
My 8150 @4.4 GHZ takes ~44K seconds without AVX.
And my old 1090T needed ~60K seconds with r409.


With each crime and every kindness we birth our future.
ID: 1211906 · Report as offensive
1 · 2 · 3 · Next

Message boards : Number crunching : APv6.01 via Lunatics 0.40 installer has errored out all APv6 tasks


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.