def solve_neutral(self, phi_g, rho_g):
# b_phi1 and b_phi2 are the boundary Hartree potential values
# of left and right sides
if self.comm_reshape:
global_rho_g0 = self.gd.collect(rho_g)
rho_g1 = self.scatter_r_distribution(global_rho_g0)
else:
rho_g1 = rho_g
# use copy() to avoid the C_contiguous=False
rho_g2 = fft2(rho_g1, None, (0, 1)).copy()
global_rho_g = self.gather_r_distribution(rho_g2)
if self.gd.comm.rank == 0:
global_rho_g.shape = (self.d1 * self.d2, self.d3)
rho_g3 = self.scatter_k_distribution(global_rho_g)
du0 = np.zeros(self.d3 - 1, dtype=complex)
du20 = np.zeros(self.d3 - 2, dtype=complex)
h2 = self.gd.h_cv[2, 2] ** 2
phi_g1 = np.zeros(rho_g3.shape, dtype=complex)
index = self.k_distribution[self.gd.comm.rank]
for phi, rho, rv2, bp1, bp2, i in zip(phi_g1, rho_g3,
self.k_vq2,
self.loc_b_phi1,
self.loc_b_phi2,
range(len(index))):
A = np.zeros(self.d3, dtype=complex) + 2 + h2 * rv2
phi = rho * np.pi * 4 * h2
phi[0] += bp1
phi[-1] += bp2
du = du0 - 1
dl = du0 - 1
du2 = du20 - 1
_gpaw.linear_solve_tridiag(self.d3, A, du, dl, du2, phi)
phi_g1[i] = phi
global_phi_g = self.gather_k_distribution(phi_g1)
if self.gd.comm.rank == 0:
global_phi_g.shape = (self.d1, self.d2, self.d3)
phi_g2 = self.scatter_r_distribution(global_phi_g, dtype=complex)
# use copy() to avoid the C_contiguous=False
phi_g3 = ifft2(phi_g2, None, (0, 1)).real.copy()
if self.comm_reshape:
global_phi_g = self.gather_r_distribution(phi_g3, dtype=float)
self.gd.distribute(global_phi_g, phi_g)
else:
phi_g[:] = phi_g3
def solve_neutral(self, phi_g, rho_g):
# b_phi1 and b_phi2 are the boundary Hartree potential values
# of left and right sides
if self.comm_reshape:
global_rho_g0 = self.gd.collect(rho_g)
rho_g1 = self.scatter_r_distribution(global_rho_g0)
else:
rho_g1 = rho_g
# use copy() to avoid the C_contiguous=False
rho_g2 = fft2(rho_g1, None, (0, 1)).copy()
global_rho_g = self.gather_r_distribution(rho_g2)
if self.gd.comm.rank == 0:
global_rho_g.shape = (self.d1 * self.d2, self.d3)
rho_g3 = self.scatter_k_distribution(global_rho_g)
du0 = np.zeros(self.d3 - 1, dtype=complex)
du20 = np.zeros(self.d3 - 2, dtype=complex)
h2 = self.gd.h_cv[2, 2] ** 2
phi_g1 = np.zeros(rho_g3.shape, dtype=complex)
index = self.k_distribution[self.gd.comm.rank]
for phi, rho, rv2, bp1, bp2, i in zip(phi_g1, rho_g3,
self.k_vq2,
self.loc_b_phi1,
self.loc_b_phi2, range(len(index))):
A = np.zeros(self.d3, dtype=complex) + 2 + h2 * rv2
phi = rho * np.pi * 4 * h2
phi[0] += bp1
phi[-1] += bp2
du = du0 - 1
dl = du0 - 1
du2 = du20 - 1
_gpaw.linear_solve_tridiag(self.d3, A, du, dl, du2, phi)
phi_g1[i] = phi
global_phi_g = self.gather_k_distribution(phi_g1)
if self.gd.comm.rank == 0:
global_phi_g.shape = (self.d1, self.d2, self.d3)
phi_g2 = self.scatter_r_distribution(global_phi_g, dtype=complex)
# use copy() to avoid the C_contiguous=False
phi_g3 = ifft2(phi_g2, None, (0, 1)).real.copy()
if self.comm_reshape:
global_phi_g = self.gather_r_distribution(phi_g3, dtype=float)
self.gd.distribute(global_phi_g, phi_g)
else:
phi_g[:] = phi_g3