Installation

Prerequisites

multiphenicsx has a few build dependencies that must be installed manually. Follow the appropriate instructions below based on how dolfinx was installed.

If dolfinx was installed via conda

conda install -c conda-forge nanobind scikit-build-core

If dolfinx was installed via apt

apt install python3-nanobind python3-scikit-build-core

If dolfinx was installed via Docker or built from source

python3 -m pip install nanobind scikit-build-core[pyproject]

Installation

multiphenicsx is available on PyPI. Use pip extras to install additional dependencies needed for the tutorials.

python3 -m pip install --no-build-isolation 'multiphenicsx[tutorials]'

Running tutorials

To run the tutorials, first clone the multiphenicsx repository. Then, ensure you check out the tag that corresponds to the version of multiphenicsx currently installed.

git clone https://github.com/multiphenics/multiphenicsx.git
cd multiphenicsx
MULTIPHENICSX_VERSION=$(python3 -c "import multiphenicsx; print(multiphenicsx.__version__)")
git checkout ${MULTIPHENICSX_VERSION}

Related resources

• Block matrix support in dolfinx, either as MatNest or monolithic matrices. In multiphenicsx we also support possible restriction of the unknowns to subdomains and/or boundaries.

• Restriction support in dolfiny relies on assembling tensors on the whole domain, and the restricting them to subdomains and/or boundaries. In multiphenicsx we directly allocate the restricted tensors, so that no unnecessary memory allocations are carried out.

• dolfinx >= 0.9.0 supports mixed dimensional assembly. The implementations of subdomain/boundary restricted variables in multiphenicsx is different to the one in dolfinx, and the two different implementations will co-exist for the foreseeable future.

• Please contact us by email if you have other related resources.