Scheduler wait error

Message boards : Number crunching : Scheduler wait error
Message board moderation

To post messages, you must log in.

AuthorMessage
MikeN

Send message
Joined: 24 Jan 11
Posts: 319
Credit: 64,719,409
RAC: 85
United Kingdom
Message 1692224 - Posted: 16 Jun 2015, 15:53:49 UTC

My GTX460 finally got an AP v7 7.05 (open cl nvidea 100) a couple of days ago for the first time in about 3 months. However, it immediately (after ca 2 seconds run time) hung with the following error in the status column on the task page (Scheduler wait: CL file build failure).

This PC and card used to crunch APs with no problem at all and it is still happily eating Einstein WU's. My driver is: NVIDIA GeForce GTX 460 (768MB) driver: 350.12 OpenCL: 1.1 and I am running Boinc 7.2.28. I am also using Lunatics anonymous app.

Any help and suggestions would be appreciated.
ID: 1692224 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1692229 - Posted: 16 Jun 2015, 16:05:40 UTC - in response to Message 1692224.  
Last modified: 16 Jun 2015, 16:06:35 UTC

My driver is: NVIDIA GeForce GTX 460 (768MB) driver: 350.12 OpenCL: 1.1


NV driver 350.12 is OpenCL 1.2

Are you using Lunatics to run or stock?

Try Rolling back to a NV driver 347.88 which still has OpenCl 1.1
ID: 1692229 · Report as offensive
MikeN

Send message
Joined: 24 Jan 11
Posts: 319
Credit: 64,719,409
RAC: 85
United Kingdom
Message 1692249 - Posted: 16 Jun 2015, 16:41:31 UTC - in response to Message 1692229.  

I am running Lunatics
ID: 1692249 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1692283 - Posted: 16 Jun 2015, 22:57:59 UTC - in response to Message 1692249.  
Last modified: 16 Jun 2015, 23:24:46 UTC

There is a problem with AP crunching under NV driver 350.12 and lunatics

If you want to run AP, there are 2 options

1) Download a previous version of NV driver ( that I gave link to) and install and crunch away

http://www.nvidia.com/download/driverResults.aspx/83080/en-us

or

2) Download the fix from Raistmer's Dropbox, decompress, move the contents to Setiathome folder and
edit the app_config.xml to reflect the new applications. Then restart Boinc and hope you did it all right

I prefer the easier option #1


Edit...Sorry forgot to post link, there it is
ID: 1692283 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1692292 - Posted: 16 Jun 2015, 23:10:03 UTC

Option #3

find your boinc seti directory and the AstroPulse_Kernels_r2793.cl or whatever your AP cl file version is.

Edit where ever there is a "bool2" to be "bool_2".

delete the according AstroPulse_Kernels_r2793.cl_* file.
(like AstroPulse_Kernels_r2793.cl_GeForceGTX780.bin_V7_TWIN_FFA_34659)

Now You're done. It works after the edit and delete.

--
An explanation: The .cl file has an user defined type bool2 that was not a reserved word or a type in .cl language when the modification was written (by me).

Nowadays the bool2 is a reserved word (a type) in the OpenCL language (v1.2). That is why the old .cl file does not work unless You do the update or the edit.
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1692292 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1692294 - Posted: 16 Jun 2015, 23:13:28 UTC

ID: 1692294 · Report as offensive
MikeN

Send message
Joined: 24 Jan 11
Posts: 319
Credit: 64,719,409
RAC: 85
United Kingdom
Message 1692567 - Posted: 17 Jun 2015, 9:36:21 UTC - in response to Message 1692292.  

Thanks everyone, went with option 3. Will have to wait until I can snag another AP to see if I got it right as the one I had errored out last night.

I did search for scheduler wait error before posting, but did not find the previous thread.
ID: 1692567 · 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 1692861 - Posted: 18 Jun 2015, 0:11:58 UTC - in response to Message 1692292.  

Option #3

find your boinc seti directory and the AstroPulse_Kernels_r2793.cl or whatever your AP cl file version is.

Edit where ever there is a "bool2" to be "bool_2".

delete the according AstroPulse_Kernels_r2793.cl_* file.
(like AstroPulse_Kernels_r2793.cl_GeForceGTX780.bin_V7_TWIN_FFA_34659)

Now You're done. It works after the edit and delete.

--
An explanation: The .cl file has an user defined type bool2 that was not a reserved word or a type in .cl language when the modification was written (by me).

Nowadays the bool2 is a reserved word (a type) in the OpenCL language (v1.2). That is why the old .cl file does not work unless You do the update or the edit.

deletion excessive. If that file would exist it meant it was already compiled. But it can't be compiled because of error. So, if it would exist all would work OK already.
ID: 1692861 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1692979 - Posted: 18 Jun 2015, 7:22:42 UTC - in response to Message 1692861.  

Option #3

find your boinc seti directory and the AstroPulse_Kernels_r2793.cl or whatever your AP cl file version is.

Edit where ever there is a "bool2" to be "bool_2".

delete the according AstroPulse_Kernels_r2793.cl_* file.
(like AstroPulse_Kernels_r2793.cl_GeForceGTX780.bin_V7_TWIN_FFA_34659)

Now You're done. It works after the edit and delete.

--
An explanation: The .cl file has an user defined type bool2 that was not a reserved word or a type in .cl language when the modification was written (by me).

Nowadays the bool2 is a reserved word (a type) in the OpenCL language (v1.2). That is why the old .cl file does not work unless You do the update or the edit.

deletion excessive. If that file would exist it meant it was already compiled. But it can't be compiled because of error. So, if it would exist all would work OK already.


Delete needed only for the previous driver version's .bin file. (For cosmetic reasons only).
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1692979 · Report as offensive
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1693036 - Posted: 18 Jun 2015, 9:32:34 UTC - in response to Message 1692229.  

My driver is: NVIDIA GeForce GTX 460 (768MB) driver: 350.12 OpenCL: 1.1

NV driver 350.12 is OpenCL 1.2

The OpenCL version shown in the GPU detection is the one that the hardware supports at maximum, not the one that the software can add. Compliance versus compatibility.

Since Nvidia stopped support development of OpenCL in favor of their own CUDA, all of its GPUs will at maximum only hardware support OpenCL 1.2, until they start development of OpenCL again (if ever).

The GTX 460 was available in the OpenCL 1.1 era, it can get drivers that support 1.2, but will only do 1.1 via hardware.
Just as the GTX 980 will only do at max 1.2, even while it's been developed and released during the OpenCL 2.0/2.1 era.
You can compare these compliancies to DirectX. You can have a DX9c compliant videocard but an OS that runs DX11. The videocard will still only support DX9c, not magically upgrade its hardware to be compliant with DX11.
ID: 1693036 · Report as offensive
MikeN

Send message
Joined: 24 Jan 11
Posts: 319
Credit: 64,719,409
RAC: 85
United Kingdom
Message 1696971 - Posted: 30 Jun 2015, 9:45:00 UTC - in response to Message 1692567.  

Thanks everyone, went with option 3. Will have to wait until I can snag another AP to see if I got it right as the one I had errored out last night.

I did search for scheduler wait error before posting, but did not find the previous thread.


Finally got some AP's and they seem to be running normally, so thanks to everyone for your advice in solving this problem.
ID: 1696971 · Report as offensive

Message boards : Number crunching : Scheduler wait error


 
©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.