Linux x86_64 CUDA app
log in

Advanced search

Message boards : Unix/Linux : Linux x86_64 CUDA app

Author Message
cronscript
Send message
Joined: 20 Aug 10
Posts: 6
Credit: 295,088,162
RAC: 0
Message 9731 - Posted: 21 Sep 2010, 16:29:23 UTC

I'm having a problem with the 2.04 cuda31 app. The first issue is that I had to remove the <file_name>libcudart.so.3</file_name> references from the app_info.xml file otherwise the application would not run, but would try to download this file.

Now that it is running, I'm seeing the following in the BOINC log file:

21-Sep-2010 07:02:56 [Collatz Conjecture] Starting collatz_1283173811_1722260_1
21-Sep-2010 07:02:56 [Collatz Conjecture] Starting task collatz_1283173811_1722260_1 using collatz version 204
21-Sep-2010 07:02:57 [Collatz Conjecture] Computation for task collatz_1283173811_1722260_1 finished
21-Sep-2010 07:02:57 [Collatz Conjecture] Output file collatz_1283173811_1722260_1_0 for task collatz_1283173811_1722260_1 absent
21-Sep-2010 07:02:57 [Collatz Conjecture] Starting collatz_1283173811_1722207_0
21-Sep-2010 07:02:57 [Collatz Conjecture] Starting task collatz_1283173811_1722207_0 using collatz version 204
21-Sep-2010 07:02:58 [Collatz Conjecture] Computation for task collatz_1283173811_1722207_0 finished
21-Sep-2010 07:02:58 [Collatz Conjecture] Output file collatz_1283173811_1722207_0_0 for task collatz_1283173811_1722207_0 absent


Any ideas as to what is causing this?

I have two machines with GTX 480's and both are doing the same thing.

Thanks,

Josh

Profile Crunch3r
Volunteer moderator
Project developer
Project tester
Avatar
Send message
Joined: 30 Jun 09
Posts: 219
Credit: 7,515,635,101
RAC: 11,492
Message 9736 - Posted: 21 Sep 2010, 19:40:55 UTC - in response to Message 9731.
Last modified: 21 Sep 2010, 19:41:15 UTC


Any ideas as to what is causing this?

I have two machines with GTX 480's and both are doing the same thing.

Thanks,

Josh

Just look at one of your results like this one here:

[url]http://boinc.thesonntags.com/collatz/result.php?resultid=51602629 [/url]

<core_client_version>6.10.56</core_client_version>
<![CDATA[
<message>
process exited with code 22 (0x16, -234)
</message>
<stderr_txt>
execv: Permission denied

</stderr_txt>
]]>

Seems as if boinc is not allowed to access the cuda libraries. You need to fix your file permissions.
____________

Team BOINC United.Join Science that matters.

cronscript
Send message
Joined: 20 Aug 10
Posts: 6
Credit: 295,088,162
RAC: 0
Message 9769 - Posted: 24 Sep 2010, 16:25:11 UTC - in response to Message 9736.

GPUGRID works fine, collatz doesn't. The permissions also seem to be ok. Any other suggestions?


[boinc@viper4 collatz_2.04_x86_64-pc-linux-gnu__cuda31]$ ldd collatz_2.04_x86_64-pc-linux-gnu__cuda31
libcuda.so.1 => /usr/lib64/libcuda.so.1 (0x00002aada2396000)
libcudart.so.3 => /usr/lib64/libcudart.so.3 (0x00002aada2d7e000)
libm.so.6 => /lib64/libm.so.6 (0x00000039e2a00000)
libc.so.6 => /lib64/libc.so.6 (0x00000039e2600000)
/lib64/ld-linux-x86-64.so.2 (0x00000039e2200000)
libpthread.so.0 => /lib64/libpthread.so.0 (0x00000039e3200000)
libz.so.1 => /usr/lib64/libz.so.1 (0x00000039e3600000)
libdl.so.2 => /lib64/libdl.so.2 (0x00000039e2e00000)
librt.so.1 => /lib64/librt.so.1 (0x00000039e6e00000)
libstdc++.so.6 => /usr/lib64/libstdc++.so.6 (0x00000039e9200000)
libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00000039e8a00000)

[boinc@viper4 collatz_2.04_x86_64-pc-linux-gnu__cuda31]$ ls -l /usr/lib64/libcuda*
lrwxrwxrwx 1 root root 14 Sep 10 16:49 /usr/lib64/libcudart.so -> libcudart.so.3
lrwxrwxrwx 1 root root 18 Sep 10 16:49 /usr/lib64/libcudart.so.3 -> libcudart.so.3.1.9
-rwxr-xr-x 1 root root 240896 Sep 10 16:49 /usr/lib64/libcudart.so.3.0.14
-rwxr-xr-x 1 root root 207432 Sep 10 16:49 /usr/lib64/libcudart.so.3.1.9
lrwxrwxrwx 1 root root 12 Sep 10 16:37 /usr/lib64/libcuda.so -> libcuda.so.1
lrwxrwxrwx 1 root root 17 Sep 10 16:37 /usr/lib64/libcuda.so.1 -> libcuda.so.256.53
-rwxr-xr-x 1 root root 8143366 Sep 10 16:37 /usr/lib64/libcuda.so.256.53

[boinc@viper4 collatz_2.04_x86_64-pc-linux-gnu__cuda31]$ ./collatz_2.04_x86_64-pc-linux-gnu__cuda31
[boinc@viper4 collatz_2.04_x86_64-pc-linux-gnu__cuda31]$

Profile Crunch3r
Volunteer moderator
Project developer
Project tester
Avatar
Send message
Joined: 30 Jun 09
Posts: 219
Credit: 7,515,635,101
RAC: 11,492
Message 9771 - Posted: 24 Sep 2010, 19:45:40 UTC - in response to Message 9769.
Last modified: 24 Sep 2010, 19:46:12 UTC

GPUGRID works fine, collatz doesn't. The permissions also seem to be ok. Any other suggestions?


[boinc@viper4 collatz_2.04_x86_64-pc-linux-gnu__cuda31]$ ldd collatz_2.04_x86_64-pc-linux-gnu__cuda31
libcuda.so.1 => /usr/lib64/libcuda.so.1 (0x00002aada2396000)
libcudart.so.3 => /usr/lib64/libcudart.so.3 (0x00002aada2d7e000)
libm.so.6 => /lib64/libm.so.6 (0x00000039e2a00000)
libc.so.6 => /lib64/libc.so.6 (0x00000039e2600000)
/lib64/ld-linux-x86-64.so.2 (0x00000039e2200000)
libpthread.so.0 => /lib64/libpthread.so.0 (0x00000039e3200000)
libz.so.1 => /usr/lib64/libz.so.1 (0x00000039e3600000)
libdl.so.2 => /lib64/libdl.so.2 (0x00000039e2e00000)
librt.so.1 => /lib64/librt.so.1 (0x00000039e6e00000)
libstdc++.so.6 => /usr/lib64/libstdc++.so.6 (0x00000039e9200000)
libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00000039e8a00000)

[boinc@viper4 collatz_2.04_x86_64-pc-linux-gnu__cuda31]$ ls -l /usr/lib64/libcuda*
lrwxrwxrwx 1 root root 14 Sep 10 16:49 /usr/lib64/libcudart.so -> libcudart.so.3
lrwxrwxrwx 1 root root 18 Sep 10 16:49 /usr/lib64/libcudart.so.3 -> libcudart.so.3.1.9
-rwxr-xr-x 1 root root 240896 Sep 10 16:49 /usr/lib64/libcudart.so.3.0.14
-rwxr-xr-x 1 root root 207432 Sep 10 16:49 /usr/lib64/libcudart.so.3.1.9
lrwxrwxrwx 1 root root 12 Sep 10 16:37 /usr/lib64/libcuda.so -> libcuda.so.1
lrwxrwxrwx 1 root root 17 Sep 10 16:37 /usr/lib64/libcuda.so.1 -> libcuda.so.256.53
-rwxr-xr-x 1 root root 8143366 Sep 10 16:37 /usr/lib64/libcuda.so.256.53

[boinc@viper4 collatz_2.04_x86_64-pc-linux-gnu__cuda31]$ ./collatz_2.04_x86_64-pc-linux-gnu__cuda31
[boinc@viper4 collatz_2.04_x86_64-pc-linux-gnu__cuda31]$



AFAIK, GPUGRID provides it's own cuda libs.

Anyway, since only root has access to your cuda libs, i doubt that the user boinc@viper4 has permission to access them.
____________

Team BOINC United.Join Science that matters.

cronscript
Send message
Joined: 20 Aug 10
Posts: 6
Credit: 295,088,162
RAC: 0
Message 9772 - Posted: 24 Sep 2010, 21:22:15 UTC - in response to Message 9771.


[boinc@viper4 collatz_2.04_x86_64-pc-linux-gnu__cuda31]$ ls -l /usr/lib64/libcuda*
lrwxrwxrwx 1 root root 14 Sep 10 16:49 /usr/lib64/libcudart.so -> libcudart.so.3
lrwxrwxrwx 1 root root 18 Sep 10 16:49 /usr/lib64/libcudart.so.3 -> libcudart.so.3.1.9
-rwxr-xr-x 1 root root 240896 Sep 10 16:49 /usr/lib64/libcudart.so.3.0.14
-rwxr-xr-x 1 root root 207432 Sep 10 16:49 /usr/lib64/libcudart.so.3.1.9
lrwxrwxrwx 1 root root 12 Sep 10 16:37 /usr/lib64/libcuda.so -> libcuda.so.1
lrwxrwxrwx 1 root root 17 Sep 10 16:37 /usr/lib64/libcuda.so.1 -> libcuda.so.256.53
-rwxr-xr-x 1 root root 8143366 Sep 10 16:37 /usr/lib64/libcuda.so.256.53

[boinc@viper4 collatz_2.04_x86_64-pc-linux-gnu__cuda31]$ ./collatz_2.04_x86_64-pc-linux-gnu__cuda31
[boinc@viper4 collatz_2.04_x86_64-pc-linux-gnu__cuda31]$



AFAIK, GPUGRID provides it's own cuda libs.

Anyway, since only root has access to your cuda libs, i doubt that the user boinc@viper4 has permission to access them.


The permissions are 755, so everyone has read and execute permission. I also have SELinux disabled, so that isn't it either.

Is there any way to get a debug log file for the collatz executable?

mickydl*
Send message
Joined: 2 Mar 10
Posts: 1
Credit: 1,099,662
RAC: 0
Message 9774 - Posted: 25 Sep 2010, 10:57:31 UTC

I had the same problem. After copying the cuda libs to the project directory (and making sure the permissions are OK) as well as to /usr/lib64 everything is working fine. I have no idea why, though.

Michael

cronscript
Send message
Joined: 20 Aug 10
Posts: 6
Credit: 295,088,162
RAC: 0
Message 9841 - Posted: 29 Sep 2010, 17:09:19 UTC - in response to Message 9774.

I just tried that as well, with no success. I don't understand what is preventing it from running, so I give up for now.

Profile mikey
Avatar
Send message
Joined: 11 Aug 09
Posts: 3242
Credit: 1,690,449,976
RAC: 5,800,577
Message 9849 - Posted: 30 Sep 2010, 11:19:41 UTC - in response to Message 9841.

I just tried that as well, with no success. I don't understand what is preventing it from running, so I give up for now.


FIRST I am NOT a Linux guy and will probably get this wrong but in another thread they talked about making sure every user on the pc is in the 'boinc users group'. I know you said your users are "755" but I have no idea what that means, so disregard if this is wrong!

cronscript
Send message
Joined: 20 Aug 10
Posts: 6
Credit: 295,088,162
RAC: 0
Message 9858 - Posted: 1 Oct 2010, 14:18:43 UTC - in response to Message 9849.

I looked at this some more and reread the README file and realized what I did incorrectly.

The part I misread is step 2.3 - I copied the entire collatz_2.04 directory to the projects directory rather than just the collatz executable. The two are named the same.

However, there is a step missing from the README file, which is the requirement to copy libcudart.so.3 to the collatz project directory as well. This file is referenced in app_info.xml for some reason, which causes boinc to check to make sure that it exists. The alternative is to remove the references to this file from app_info.xml.


ironcold
Send message
Joined: 12 Jul 09
Posts: 7
Credit: 11,787,126
RAC: 0
Message 9859 - Posted: 1 Oct 2010, 17:26:40 UTC - in response to Message 9774.

I had the same problem. After copying the cuda libs to the project directory (and making sure the permissions are OK) as well as to /usr/lib64 everything is working fine. I have no idea why, though.

Michael

I'm having a problem with the 2.04 cuda31 app. The first issue is that I had to remove the <file_name>libcudart.so.3</file_name> references from the app_info.xml file otherwise the application would not run


Same problem here with SUSE11.3, GTS450, NVIDIA-Linux-x86_64-260.19.04. I also had to remove the <file_ref> <file_name>libcudart.so.3</file_name> </file_ref> to make it work.

Gary Craig
Send message
Joined: 30 Jan 10
Posts: 6
Credit: 3,626,547
RAC: 0
Message 12404 - Posted: 29 May 2011, 8:45:00 UTC

I made a symlink from the project directory to where I copied libcudart.so.3 in /usr/lib64. That cleared things up. This is with Ubuntu 10.10.

--Gary


Post to thread

Message boards : Unix/Linux : Linux x86_64 CUDA app


Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.