#include <polar_poisson_solver.hpp>
◆ PolarPoissonSolver()
- Parameters
-
| geom | annulus (r, theta), domain.nx() = nr radial cells, ny() = ntheta azimuthal. |
| ba | BoxArray with ONE box covering the whole annulus (single-rank). |
| bc | radial BC: xlo/xhi (Dirichlet -> xlo_val/xhi_val; Foextrap -> homogeneous Neumann). theta (ylo/yhi) is treated as PERIODIC regardless of the value (FFT). |
◆ geom()
◆ phi()
| MultiFab & pops::PolarPoissonSolver::phi |
( |
| ) |
|
|
inline |
◆ residual()
| Real pops::PolarPoissonSolver::residual |
( |
| ) |
|
|
inline |
Discrete residual ||L_h phi - rhs|| (inf norm), reduced over the ranks.
L_h is the EXACT operator that solve() inverts: FFT in theta (mode m, spectral eigenvalue -k(m)^2/r^2) + conservative order-2 radial stencil, folded radial BC (Dirichlet: constant value = m=0 part; Neumann: homogeneous reflection). We evaluate it MODE BY MODE in Fourier space (where the theta diagonalization is exact), so the residual reflects ONLY the quality of the solve (and not the spectral gap vs finite differences of a physical theta stencil). For this DIRECT solver, ~round-off. The m=0 double-Neumann mode is EXCLUDED (its row 0 is pinned = gauge: trivial identity equation).
◆ rhs()
| MultiFab & pops::PolarPoissonSolver::rhs |
( |
| ) |
|
|
inline |
◆ solve()
| void pops::PolarPoissonSolver::solve |
( |
| ) |
|
|
inline |
Solves (1/r) d_r(r d_r phi) + (1/r^2) d_theta^2 phi = rhs in place, by FFT-in-theta then a tridiagonal solve (Thomas) per azimuthal mode.
phi() holds the solution after the call.
The documentation for this class was generated from the following file:
Collaboration diagram for pops::PolarPoissonSolver:
1.9.8