This may be a simple question, but I cannot find answer.
is there some library to map multi-d array to the CUDA grid/block/thread hierarchy, so developer can manipulate the the array index with little efforts (not the blockIdx/threadIdx/… computation each time)? something like in MPI_Cart_create. If no, how good and beneficial for developers to have such library for CUDA?
also another simpler one, can I create a multi-d array (like int myarray ) in device code?
thanks very much!