Nvcc fatal : Visual Studio configuration file '(null)'

Byte-Welt - Projekte / Projects JCuda
1

Hey

I get the following error when attemping to run the „Matrix Inversion“ sample

Loading kernels from GPUadjustRow_kernel.cu
26-Dec-2010 23:58:32 jcuda.utils.KernelLauncher prepareCubinFile
SEVERE: errorMessage:
nvcc fatal : Visual Studio configuration file ‚(null)‘ could not be found for installation at ‚C:\IDE\VS2008E\VC\bin/…/…‘

26-Dec-2010 23:58:32 jcuda.utils.KernelLauncher prepareCubinFile
SEVERE: outputMessage:

Exception in thread „main“ jcuda.CudaException: Could not create .cubin file: nvcc fatal : Visual Studio configuration file ‚(null)‘ could not be found for installation at ‚C:\IDE\VS2008E\VC\bin/…/…‘

at jcuda.utils.KernelLauncher.prepareCubinFile(KernelLauncher.java:1145)
at jcuda.utils.KernelLauncher.create(KernelLauncher.java:385)
at jcuda.utils.KernelLauncher.create(KernelLauncher.java:321)
at com.pbd.cuda.sample.matrix.MatrixInvert.init(MatrixInvert.java:62)
at com.pbd.cuda.sample.matrix.MatrixInvert.invert(MatrixInvert.java:114)
at com.pbd.cuda.sample.matrix.MatrixInvertSample.main(MatrixInvertSample.java:32)

I’m running CUDA 32x/Java 32x.

Running a C++ example I have using NMAKE works fine, so I’m slightly stumped at the issue.

I had run into this very same „config file ‚(null)‘“ thing before, but that’s because I was trying to use CUDA 64x with a VS 32x (or so I believe that was the cause as changing CUDA to 32x seemed to fix it)

If anyone has any idea how to solve this I would very much appreciate it.

Thanks for all the work on the library. :slight_smile:

Edit: on the upside, I have JCudaRuntimeGLSample3.java working… thank blod for 64x JOGL

2

Hello,

I’ve never encountered this error message. It might be a wrong version of VS installed, maybe some 32/64 bit issue? However, a websearch on the error message brings a lot of (CUDA-specific) results, among them one answer on StackOverflow which states to have the problem solved. You may want to give it a try…

BTW: The GL samples are slightly out-dated (as stated on the samples page). The GL interoperability has changed slightly after CUDA 3.0, and the samples have not yet been updated. Additionally, the latest version of JOGL has changed as well, but this mainly consists of some renamings. The task to update these samples has nearly reached the top of my TODO-list :wink:

bye
Marco