Incorrect texture address mode in convolutionFFT2D sample

According to https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#texture-memory

cudaAddressModeWrap and cudaAddressModeMirror are only supported for normalized texture coordinates

In "convolutionFFT2D.cu" the texture address mode is set as "cudaAddressModeWrap" and coordinates isn't set as normalized. For example:

extern "C" void padKernel(float *d_Dst, float *d_Src, int fftH, int fftW,
int kernelH, int kernelW, int kernelY, int kernelX) {
...

texDescr.normalizedCoords = false;
texDescr.filterMode = cudaFilterModeLinear;
texDescr.addressMode[0] = cudaAddressModeWrap; // This address mode is only supported for normalized texture coordinates.
texDescr.readMode = cudaReadModeElementType;
...
}

Start with convolutionFFT2D.cu and inspect the padKernel texture descriptor, especially normalizedCoords and addressMode[0]. Compare the configuration with the CUDA texture-memory documentation cited in the issue. Done means the sample no longer uses an unsupported address-mode and remains consistent with its coordinate configuration.