CUDA unknown error

Question

I'm trying to run mainSift.cpp from CudaSift on a Nvidia Tesla M2090. First of all, as explained in this question, I had to change from sm_35 to sm_20 the CMakeLists.txt .

Unfortunatley now this error is returned:

checkMsg() CUDA error: LaplaceMulti() execution failed
 in file </ghome/rzhengac/Downloads/CudaSift/cudaSiftH.cu>, line 318 : unknown error.

And this is the LaplaceMulti code:

double LaplaceMulti(cudaTextureObject_t texObj, CudaImage *results, float baseBlur, float diffScale, float initBlur)
{
  float kernel[12*16];
  float scale = baseBlur;
  for (int i=0;i<NUM_SCALES+3;i++) {
    float kernelSum = 0.0f;
    float var = scale*scale - initBlur*initBlur;
    for (int j=-LAPLACE_R;j<=LAPLACE_R;j++) {
      kernel[16*i+j+LAPLACE_R] = (float)expf(-(double)j*j/2.0/var);
      kernelSum += kernel[16*i+j+LAPLACE_R]; 
    }
    for (int j=-LAPLACE_R;j<=LAPLACE_R;j++) 
      kernel[16*i+j+LAPLACE_R] /= kernelSum;  
    scale *= diffScale;
  }
  safeCall(cudaMemcpyToSymbol(d_Kernel2, kernel, 12*16*sizeof(float)));
  int width = results[0].width;
  int pitch = results[0].pitch;
  int height = results[0].height;
  dim3 blocks(iDivUp(width+2*LAPLACE_R, LAPLACE_W), height);
  dim3 threads(LAPLACE_W+2*LAPLACE_R, LAPLACE_S); 
  LaplaceMulti<<<blocks, threads>>>(texObj, results[0].d_data, width, pitch, height);
  checkMsg("LaplaceMulti() execution failed\n");
  return 0.0; 
}

I've read already this question that seems somewhat similar, but I don't understand what the solution means or how to use it for my problem.

Why does the error occur?

Answer 1

The error occurs because you are running code which has features which are not supported on your GPU (texture objects). I am a little surprised that the compiler doesn't generate an error during compilation, but that is another question.

There is no solution except to use supported hardware, or to rewrite the code.

[This answer assembled from comments and added as a community wiki entry to get this answer off the unanswered list for the CUDA tag]