Use OpenACC to implement a Jacobi iteration code that does the computation entirely on the GPU device.
Starting from a serial code that implements a Jacobi iteration (provided in
file jacobi(.c|.F90), modify the code is such a way that the computation is
done entirely on the device without any additional data transfers taking place
between the host and the device.