I would like to allocate an array of, say, 14 x 13 floating point values and bind a texture to it.
If I use cudaMallocPitch to allocate the array, I can bind a texture to the array using cudaBindTexture2D and all operations on the texture work just fine (e.g., interpolation)
If I use cudaMalloc (that is the width of the array is not pitched to 64 bytes), I can still bind a texture using cudaBindTexture2D but most operations on the texture do not provide the correct result.
According to this reference
allocating with cudaMalloc should not be a problem. Any idea?
Thanks for your help,