-
Notifications
You must be signed in to change notification settings - Fork 101
Implicit methods
- Jed Brown
- Matteo Parsani
- Lulu Liu
- Implementation of "some" implicit TS schemes for the solution of stiff geophysical problem
We are maybe aiming too high but one should always do that :-)
- Nonlinear geometric multigrid for low-Mach transient problems (finite time step).
- Multi-stage smoothers (aka explicit Runge-Kutta, similar to polynomial smoothers).
- Constructing of diagonal preconditioner using wave decomposition.
- Possible use of matrix-based Gauss-Seidel or ILU smoothers.
github repository: https://github.com/mparsani/pyclaw/tree/impl-hpc3
4 February 2012
-
PETSc can use explicit RK schemes in 3S* low-storage form as smoother for MG. ERK62 has been tested for the lid-driven cavity flow available in PETSc. Its performance are not impressive but it has not been design for damping high-frequencies. In the next weeks we will provide some optimized smoothers. Van Leer smoothers are also available in PETSc now.
-
interpolation of PyClaw aux array available in PETSc. This is required for MG. Piece-wise constant interpolation is done. This is consistent with the well-balancing setup of sharpclaw solver (drop high-order contribution and move down to second-order accurate scheme)