print("Now we do a jacobi step using sparse matrix algorithms:") print("But just before one needs to fill the rhs") print("rhs at the beginning: \n", low_level.rhs) mg_problem.fill_rhs(low_level) print("rhs after filling it: \n", low_level.rhs) # laplace_stencil.modify_rhs(low_level) print("rhs after modification: \n", low_level.rhs) jacobi_matrix.relax() print(low_level.arr) print("Now we check if the more general SplitSmootherClass :") mg_problem.fill_rhs(low_level) low_level.mid[:] = 105.0 low_level.pad() low_jacobi_smoother.relax() print(low_level.arr) print("===== Restriction and Interpolation =====") # generate restriction stencil rst_inject = RestrictionStencilPure(np.asarray([1.0]), 2) rst_fw = RestrictionStencilPure(np.asarray([0.25, 0.5, 0.25]), 2) # try it with just some simple arrays x_in = np.arange(9) ** 2 x_out = np.zeros(5) print("test injection,\n x_in :", x_in) print(" x_out :", x_out) rst_inject.eval(x_in, x_out) print("inject,\n x_out :", x_out) # full weighting restriction needs another interpretation because # the stencil needs also the values on the boundary, this men
print("level.arr after direct solution\n", level.mid) print("test of the solution Ax=b by convolve\n ", laplace_stencil.eval_convolve(level.mid, "same")) rhs_test = np.zeros(level.rhs.shape) laplace_stencil.eval_sparse(level.mid, rhs_test) print("test of the solution Ax=b by sparse matrix application \n", rhs_test) print("==== SplitSmoother ====") omega = 2.0/3.0 l_plus = np.asarray([[0, 0, 0], [0, -4.0/omega, 0], [0, 0, 0]]) l_minus = np.asarray([[0, 1.0, 0], [1.0, -4.0*(1.0 - 1.0/omega), 1.0], [0., 1., 0.]]) jacobi_smoother = SplitSmoother(l_plus, l_minus, level) level.mid[:] = 0 jacobi_smoother.relax() print("level.arr after one jacobi step with modified rhs\n", level.arr) level.mid[:] = 0. level.modified_rhs = False level.rhs[:] = 0. jacobi_smoother = SplitSmoother(l_plus, l_minus, level) jacobi_smoother.relax() print("level.arr after one jacobi step with unmodified rhs\n", level.arr) # For the test of the level transitioning we define 3 levels # with different roles but the same borders n_jacobi_pre = 5 n_jacobi_post = 5 borders = np.ones((2, 2)) top_level = MultigridLevel2D((259, 259), mg_problem=mg_problem,