Figure 6: Matrix vector multiplication by cross-combining storage.