NotebooksΒΆ
Two Jupyter notebooks work through complete problems on a graded 2D mesh, building up one step at a time and plotting as they go.
Notebook |
What it builds |
|---|---|
Graded mesh, a Mittag-Leffler warm-up, the 2D Caputo problem, a manufactured solution, separating mode and timestep error, and graded timesteps |
|
The spectral operator on a known eigenfunction, the sinc quadrature knob and its discretization floor, solving \((-\Delta_D)^s u=1\), why grading pays, and coupling back to a Caputo derivative |
Run them in an activated Firedrake environment. Beyond Yonderdrake itself, they
need only matplotlib and mpmath. They are stored without outputs, so every
figure is produced by the run in front of you.
A Jupyter kernel is a single process, so the notebooks run serially however the
environment was launched. They show each step of the calculation. For larger
problems, use mpiexec -n N python your_script.py, as in demos/.
Driving MPI from Jupyter via ipyparallel is currently untested.