Table .

Algorithm VIII. Implementation of function fwd_pthread (on host).

Input:npth, ${\tilde{\mathbf{p}}}_0$, αpth, ${\tilde{\mathbf{u}}}_{\mathrm{pth}}^{\prime }$

Output:${\tilde{\mathbf{u}}}_{\mathrm{pth}}^{\prime }$

1: ${\mathrm{C}\_\tilde{\mathbf{p}}}_0\leftarrow {\tilde{\mathbf{p}}}_0$ {Copy from host to constant memory}

2: fornx = 0 toNx/Npth − 1 do

3:   x = (nx + npthNx/Npth)Δs + xmin

4:   forny = 0 toNy − 1 do

5:      y = nyΔs + ymin

6:      $\mathrm{C}\_{\mathit{\alpha }}_{\mathrm{pth}}\leftarrow {\mathit{\alpha }}_{\mathrm{pth}}\left[n_x\right]\left[n_y\right][:]$ {Copy from host to constant

memory}

7:      K_fwdsph  ⟨⟨⟨ (Nv, Nr), (Nf, 1, 1) ⟩⟩⟩ (x, y, D_${\tilde{\mathbf{u}}}_{\mathrm{pth}}^{\prime }$)

8:   endfor

9: endfor

10: ${\tilde{\mathbf{u}}}_{\mathrm{pth}}^{\prime }\leftarrow \mathrm{D}\_{\tilde{\mathbf{u}}}_{\mathrm{pth}}^{\prime }$ {Copy from global memory to host}