FEniCSx Setup

FEniCSx is required for models involving complex geometries, such as flow around obstacles. This guide covers installation and troubleshooting.

Why FEniCSx?

Spectral methods work well for periodic boundary conditions on rectangular domains. For problems with:

Obstacles (cylinders, airfoils)
Non-rectangular domains
Dirichlet/Neumann boundaries

we use the finite element method via FEniCSx.

Installation

Using the Setup Script

The simplest approach:

cd pdeforge
./setup_fenicsx_env.sh

This creates a conda environment pdeforge-fenicsx with all dependencies.

To specify a custom name:

./setup_fenicsx_env.sh my-env-name

Manual Installation

# Create environment
conda create -n pdeforge-fenicsx python=3.11
conda activate pdeforge-fenicsx

# Install FEniCSx and dependencies
conda install -c conda-forge \
    fenics-dolfinx \
    mpich \
    petsc4py \
    gmsh \
    python-gmsh \
    pyvista

# Install PDEForge
pip install -e .

Verifying the Installation

# Test FEniCSx
import dolfinx
print(f"FEniCSx version: {dolfinx.__version__}")

# Test PDEForge with FEniCSx models
from pdeforge import list_models
models = list_models()
print(f"Available models: {models}")
assert "cylinder_flow_2d" in models

Available FEniCSx Models

Model	Description
`cylinder_flow_2d`	Steady flow around a circular cylinder
`cylinder_flow_2d_unsteady`	Time-dependent flow with vortex shedding

Troubleshooting

MPI Library Conflicts

If you see errors like Library not loaded: @rpath/libmpi.dylib:

Remove the conda environment completely
Reinstall from scratch using the setup script
Avoid mixing pip and conda installations for MPI-related packages

gmsh Issues

Install gmsh only through conda-forge:

conda install -c conda-forge gmsh python-gmsh

Do not use pip install gmsh in a FEniCSx environment.

Memory Issues

FEniCSx models can be memory-intensive. For large meshes:

Reduce mesh resolution via _mesh_resolution parameter
Reduce output grid resolution
Generate samples in smaller batches

Jupyter Integration

Register the environment as a Jupyter kernel:

conda activate pdeforge-fenicsx
python -m ipykernel install --user --name pdeforge-fenicsx

Then select "pdeforge-fenicsx" in Jupyter.