CUDA 2.0 - cuFFT fails

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi all,

I just started using CUDA and am playing around with the available samples.
My aim is right now to get the sample CUDA FFT code running.
The compilation succeeds, but the processing however fails with this message:

cufft: ERROR: /Volumes/Builds/nv/build/rel/gpgpu/toolkit/r2.0/cufft/src/config.cu, line 106
cufft: ERROR: CUFFT_INTERNAL_ERROR
cufft: ERROR: /Volumes/Builds/nv/build/rel/gpgpu/toolkit/r2.0/cufft/src/cufft.cu, line 112
cufft: ERROR: CUFFT_INVALID_VALUE
cufft: ERROR: /Volumes/Builds/nv/build/rel/gpgpu/toolkit/r2.0/cufft/src/cufft.cu, line 112
cufft: ERROR: CUFFT_INVALID_VALUE
cufft: ERROR: /Volumes/Builds/nv/build/rel/gpgpu/toolkit/r2.0/cufft/src/cufft.cu, line 94
cufft: ERROR: CUFFT_INVALID_PLAN

The path you see in the message seems to be hard coded into the libraries:
$ grep -R -n ‘/Volumes/Builds/nv/build/’ *
Binary file bin/cudafe matches
Binary file bin/cudafe++ matches
Binary file lib/libcublas.dylib matches
Binary file lib/libcublasemu.dylib matches
Binary file lib/libcuda.dylib matches
Binary file lib/libcufft.dylib matches
Binary file lib/libcufftemu.dylib matches

Moreover, the path does not exist on my machine and must be a relict from the installation package?
Does anybody have a clue how to solve that?

Thank you in advance for your help,
Thomas

code:

#include <cuda.h>
#include <cufft.h>
//#include
//#include <cuComplex.h>
#include <device_launch_parameters.h>
#include <cuda_runtime_api.h>

#include

#define NX 256
#define NY 128

int main(int argc, char** argv){

cufftHandle plan;
cufftComplex *idata, *odata;
cudaMalloc((void**)&idata, sizeof(cufftComplex)*NX*NY);
cudaMalloc((void**)&odata, sizeof(cufftComplex)*NX*NY);
/* Create a 2D FFT plan. */
cufftPlan2d(&plan, NX, NY, CUFFT_C2C);
/* Use the CUFFT plan to transform the signal out of place. */
cufftExecC2C(plan, idata, odata, CUFFT_FORWARD);
/* Note: idata != odata indicates an out-of-place transformation
         to CUFFT at execution time
 */
/* Inverse transform the signal in place */
cufftExecC2C(plan, odata, odata, CUFFT_INVERSE);
/* Destroy the CUFFT plan. */
cufftDestroy(plan);
cudaFree(idata); cudaFree(odata);

return 0;

}

2

Can you run the deviceQuery example?

I think you did not install the toolkit properly ( go to custom installation and check that the Kext is selected, it should ask you to reboot the system if the installation was successful)

3

Thank you, that was exactly the problem.

Cheers,
Thomas

4

Thank you, that was exactly the problem.

Cheers,
Thomas

5

I’ve encountered the error on Mac OSX which doesn’t have that custom installation option. Any suggestion of what to try to suppress such messages?
-jm