Now disabled @ SETI@home: VLAR WUs not to GPUs


log in

Advanced search

Message boards : Number crunching : Now disabled @ SETI@home: VLAR WUs not to GPUs

Previous · 1 · 2
Author Message
ClaggyProject donor
Volunteer tester
Send message
Joined: 5 Jul 99
Posts: 4068
Credit: 32,909,554
RAC: 7,960
United Kingdom
Message 1019817 - Posted: 26 Jul 2010, 21:38:07 UTC - in response to Message 1019816.
Last modified: 26 Jul 2010, 21:41:53 UTC

The WUs have endings of .vlar_0 and .vlar_1 and are DLed for '608 cuda' (app_info.xml).

You're right, this .VLAR Wu has been sent to the GPU, maybe because it's a resend? (Edit: i now see you have others sent to the GPU, that aren't resends)

http://setiathome.berkeley.edu/workunit.php?wuid=636108071

Claggy

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8465
Credit: 48,987,573
RAC: 73,991
United Kingdom
Message 1019823 - Posted: 26 Jul 2010, 21:42:40 UTC - in response to Message 1019817.

The WUs have endings of .vlar_0 and .vlar_1 and are DLed for '608 cuda' (app_info.xml).

You're right, this .VLAR Wu has been sent to the GPU, maybe because it's a resend?

http://setiathome.berkeley.edu/workunit.php?wuid=636108071

Claggy

Interesting. This sounds like a bug: more evidence please, guys.

I wonder whether the new scheduler rules for "Don't send .vlar to CUDA" have a hidden exception "except for anonymous platform" - wouldn't be the first time that's happened with DA's code. We're still waiting for debugged server-DCF for anonymous platform, remember.

Profile [seti.international] Dirk SadowskiProject donor
Volunteer tester
Avatar
Send message
Joined: 6 Apr 07
Posts: 7062
Credit: 60,019,049
RAC: 21,424
Germany
Message 1019825 - Posted: 26 Jul 2010, 21:44:35 UTC - in response to Message 1019817.
Last modified: 26 Jul 2010, 21:50:06 UTC

The WUs have endings of .vlar_0 and .vlar_1 and are DLed for '608 cuda' (app_info.xml).

You're right, this .VLAR Wu has been sent to the GPU, maybe because it's a resend?

http://setiathome.berkeley.edu/workunit.php?wuid=636108071

Claggy


This is a .vlar_2 WU.
Strange, I have this WU not in my BOINC. So I guess it's a ghost WU.

I have current 41 .vlar_0 and .vlar_1 WUs for GPU in my BOINC (So this are new ones).
So later I need to use the ReSchedule tool for to protect my GPU for VLAR WUs.
____________
BR



>Das Deutsche Cafe. The German Cafe.<

Profile John Neale
Volunteer tester
Avatar
Send message
Joined: 16 Mar 00
Posts: 507
Credit: 3,072,605
RAC: 2,312
South Africa
Message 1019828 - Posted: 26 Jul 2010, 21:50:55 UTC - in response to Message 1019825.

Sutaru, you run optimised ("anonymous") apps at SETI Beta. Did you observe this behaviour there?
____________

Profile [seti.international] Dirk SadowskiProject donor
Volunteer tester
Avatar
Send message
Joined: 6 Apr 07
Posts: 7062
Credit: 60,019,049
RAC: 21,424
Germany
Message 1019831 - Posted: 26 Jul 2010, 21:58:11 UTC - in response to Message 1019828.

Sutaru, you run optimised ("anonymous") apps at SETI Beta. Did you observe this behaviour there?


At SETI@home Beta Test the '.vlar WUs not to GPU' ran well with app_info.xml file.
But there I have 609 and cuda23 entries in my app_info.xml file.

____________
BR



>Das Deutsche Cafe. The German Cafe.<

JohnDKProject donor
Volunteer tester
Avatar
Send message
Joined: 28 May 00
Posts: 840
Credit: 43,096,700
RAC: 75,253
Denmark
Message 1019832 - Posted: 26 Jul 2010, 21:59:40 UTC

I've not been getting any .vlars for my GPU, has anybody else (except Sutaru)?

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8465
Credit: 48,987,573
RAC: 73,991
United Kingdom
Message 1019835 - Posted: 26 Jul 2010, 22:06:24 UTC - in response to Message 1019832.

I've not been getting any .vlars for my GPU, has anybody else (except Sutaru)?

I was away for the weekend, but I've just looked through my BoincView logs: no sign of any .vlar WU being downloaded on any of my CUDA boxes (they only fetch GPU work from SETI - CPUs are kept busy by other projects).

Doesn't mean that the project server code hasn't changed today, of course.

Profile perryjay
Volunteer tester
Avatar
Send message
Joined: 20 Aug 02
Posts: 3377
Credit: 15,331,724
RAC: 11,537
United States
Message 1019841 - Posted: 26 Jul 2010, 22:25:01 UTC

I've only got 14 of the VLAR marked WUs but they all were assigned to my CPUs. I've also got 2 non-marked VLARs but they also went to my CPUs. My reschedule v1.9 is starting to feel neglected. :-)
____________


PROUD MEMBER OF Team Starfire World BOINC

Profile Geek@PlayProject donor
Volunteer tester
Avatar
Send message
Joined: 31 Jul 01
Posts: 2466
Credit: 85,774,100
RAC: 27,777
United States
Message 1019847 - Posted: 26 Jul 2010, 22:33:57 UTC
Last modified: 26 Jul 2010, 22:44:52 UTC

All my vlar's are properly assigned to the cpu's. I have several that are _2 reissues and still to cpu's.

http://setiathome.berkeley.edu/workunit.php?wuid=636077927
http://setiathome.berkeley.edu/workunit.php?wuid=635619180

Both to cpu's as required.
____________
Boinc....Boinc....Boinc....Boinc....

Harri Liljeroos
Avatar
Send message
Joined: 29 May 99
Posts: 46
Credit: 18,832,645
RAC: 11,610
Finland
Message 1020094 - Posted: 27 Jul 2010, 13:41:29 UTC

I've got four of these too on my host. All .vlar_0 and .vlar_1. And I also have app_info for 6.08 and cuda23.

This one 01jn10ab.19594.7429.16.10.176.vlar was sent also to a standard 6.09 application.
____________

Profile MiepProject donor
Volunteer moderator
Avatar
Send message
Joined: 23 Jul 99
Posts: 2411
Credit: 351,996
RAC: 0
Message 1020109 - Posted: 27 Jul 2010, 14:28:40 UTC

ok, so to conclude, while most people don't see vlar marked tasks on their GPU a few seem to have been slipping through the net.

I checked a few of Harri's vlar to GPU WUs and in most cases both tasks have been sent to a GPU.

So some vlar issued to stock 6.09 and some to ap - what are you running Sutaru? 6.09 but in app_info?

And Harri's wingman to the 6.09 is running V12.

Doesn't look like much of a pattern.
____________
Carola
-------
I'm multilingual - I can misunderstand people in several languages!

Profile [seti.international] Dirk SadowskiProject donor
Volunteer tester
Avatar
Send message
Joined: 6 Apr 07
Posts: 7062
Credit: 60,019,049
RAC: 21,424
Germany
Message 1020114 - Posted: 27 Jul 2010, 14:35:20 UTC - in response to Message 1020109.
Last modified: 27 Jul 2010, 14:40:39 UTC

At SETT@home Beta Test I had 609_cuda23 entries in my app_info.xml file, with stock 6.09 CUDA app. And everything fine.

Here at SETI@home I have 608_cuda entries in my app_info.xml file, with stock 6.09 CUDA app. And my BOINC get .vlar_0 and .vlar_1 WUs for GPU.


EDIT: BTW. Harri, you need to delete the '/' at the end of your URL for well redirection.
____________
BR



>Das Deutsche Cafe. The German Cafe.<

Profile Geek@PlayProject donor
Volunteer tester
Avatar
Send message
Joined: 31 Jul 01
Posts: 2466
Credit: 85,774,100
RAC: 27,777
United States
Message 1020119 - Posted: 27 Jul 2010, 14:44:16 UTC

Is it possible that some are using the redistribute and inadvertantly move VLAR from CPU to GPU while trying to keep GPU's busy.
____________
Boinc....Boinc....Boinc....Boinc....

Harri Liljeroos
Avatar
Send message
Joined: 29 May 99
Posts: 46
Credit: 18,832,645
RAC: 11,610
Finland
Message 1020162 - Posted: 27 Jul 2010, 16:13:18 UTC

@Sutaru

Thanks for pointing that extra "/" out, cut and paste fault. Unfortunately can't edit it anymore.

@Miep

Actually mine is the V12 and my wingman has the 6.09. And he got it first.
____________

Profile MiepProject donor
Volunteer moderator
Avatar
Send message
Joined: 23 Jul 99
Posts: 2411
Credit: 351,996
RAC: 0
Message 1020176 - Posted: 27 Jul 2010, 21:38:39 UTC - in response to Message 1020119.

Is it possible that some are using the redistribute and inadvertantly move VLAR from CPU to GPU while trying to keep GPU's busy.


No, website shows what's been issued to, you need to check the task itself to see where it's been processed (after is has come in).

Sorry Harry :) Misread your post as yours being the 6.09, whilst you were actually pointing out that the second task had gone to GPU as well and to stock on top.

____________
Carola
-------
I'm multilingual - I can misunderstand people in several languages!

Profile MiepProject donor
Volunteer moderator
Avatar
Send message
Joined: 23 Jul 99
Posts: 2411
Credit: 351,996
RAC: 0
Message 1020491 - Posted: 29 Jul 2010, 0:59:41 UTC

Problem found - see Matt's post.

Can't expect vlar marked tasks not to end up on the GPU if the discriminating code is not turned on ;)
____________
Carola
-------
I'm multilingual - I can misunderstand people in several languages!

Profile [seti.international] Dirk SadowskiProject donor
Volunteer tester
Avatar
Send message
Joined: 6 Apr 07
Posts: 7062
Credit: 60,019,049
RAC: 21,424
Germany
Message 1020554 - Posted: 29 Jul 2010, 5:49:20 UTC
Last modified: 29 Jul 2010, 5:56:15 UTC

Thanks for the hint.

Ohh.. strange..
At SETI@home Beta Test it worked very well..


EDIT of the title of this thread from 'Now enabled @ ..' to 'Now disabled @ ..'.
____________
BR



>Das Deutsche Cafe. The German Cafe.<

Profile Donald L. JohnsonProject donor
Avatar
Send message
Joined: 5 Aug 02
Posts: 6133
Credit: 676,332
RAC: 1,129
United States
Message 1020568 - Posted: 29 Jul 2010, 7:15:56 UTC - in response to Message 1020491.

Problem found - see Matt's post.

Can't expect vlar marked tasks not to end up on the GPU if the discriminating code is not turned on ;)


As I read Matt's note, the VLAR change was only turned off for a couple of hours on Monday, so he fixed it after he became aware of it. Still, it was long enough that some folks got VLARs sent to their GPUs. And we will still have resends of VLARs split before the changes went in, probably for a couple of months.

Previous · 1 · 2

Message boards : Number crunching : Now disabled @ SETI@home: VLAR WUs not to GPUs

Copyright © 2014 University of California