Cache partitioning for sparse matrix-vector multiplication on the A64FX