Memory copy problem in CUDA 3.0 but not in 2.3

Accelerated Computing CUDA CUDA Programming and Performance
#1

I seem to have a strange problem, when I run my program it works fine for some sizes (I’m working on 3D data, i.e. volumes) but for others I get “invalid argument” when I try to run a “cudaMemcpy3D” from global memory to an array bound to a 3D texture. I don’t remember having these problems in CUDA 2.3. Does anyone know if something has changed for texture memory and cudaMemcpy3D from CUDA 2.3 to 3.0 ?