Benchmark option in oclNbody does not work extension "NV-GLX" missing on display

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello,

I want to run oclNbody in our Linux GPU batch system. I use:

oclNbody --benchmark --n=10000 --noprompt

However, it seems that the benchmark-option does not work, as I get the error:

Calling InitGL...

Xlib:  extension "NV-GLX" missing on display "localhost:10.0".

freeglut (oclNbody): Unable to create direct context rendering for window 'OpenCL for GPU Nbody Demo'

This may hurt performance.

!!! Error # -9999 (Unspecified Error) at line 355 , in file src/oclNbody.cpp !!!

I scanned oclNbody.cpp file and it really looks like the “benchmark” option is never read and used.

Is that a fault in the SDK example?

Or do I have to change anything?

Cheers, Sandra

System details:

Linux RedHat 4.1

CUDA Toolkit 3.1

Nvidia SDK 3.1

Tesla C1060

Driver version: 256.53

2

Hello,

I want to run oclNbody in our Linux GPU batch system. I use:

oclNbody --benchmark --n=10000 --noprompt

However, it seems that the benchmark-option does not work, as I get the error:

Calling InitGL...

Xlib:  extension "NV-GLX" missing on display "localhost:10.0".

freeglut (oclNbody): Unable to create direct context rendering for window 'OpenCL for GPU Nbody Demo'

This may hurt performance.

!!! Error # -9999 (Unspecified Error) at line 355 , in file src/oclNbody.cpp !!!

I scanned oclNbody.cpp file and it really looks like the “benchmark” option is never read and used.

Is that a fault in the SDK example?

Or do I have to change anything?

Cheers, Sandra

System details:

Linux RedHat 4.1

CUDA Toolkit 3.1

Nvidia SDK 3.1

Tesla C1060

Driver version: 256.53

3

The problem is that oclNbody (or rather the shrFindFilePath function in $NVIDIA_GPU_SDK/shared/src/shrUtils.cpp) is not able to find the source for its kernels. The easiest way to fix it is to run the sample from it’s own directory ($NVIDIA_GPU_SDK/OpenCL/bin/linux/release in my case).

Hope that helps.

4

The problem is that oclNbody (or rather the shrFindFilePath function in $NVIDIA_GPU_SDK/shared/src/shrUtils.cpp) is not able to find the source for its kernels. The easiest way to fix it is to run the sample from it’s own directory ($NVIDIA_GPU_SDK/OpenCL/bin/linux/release in my case).

Hope that helps.