cbc.block 3.0.0

"Block algebra for Fenics/DOLFIN",

cbc.block is a python library for block operations in DOLFIN (http://fenicsproject.org). The headline features are:

• Block operators may be defined from standard DOLFIN matrices and vectors:

  A = assemble(...); B = assemble(...); # etc
  AA = block_mat([[A,B], [C,D]])
  

• Preconditioners, inverses, and inner solvers are supported:

  AAprec = AA.scheme('gauss-seidel', inverse=ML)
  

• A good selection of iterative solvers:

  AAinv = SymmLQ(AA, precond=AAprec)
  x = AAinv*b
  

• Matrix algebra is supported both through composition of operators... :

  S = C*ILU(A)*B-D
  Sprec = ConjGrad(S)
  

  ...and through explicit matrix calculation via PyTrilinos:

  S = C*InvDiag(A)*B-D
  Sprec = ML(collapse(S))
  

There is no real documentation apart from the python doc-strings, but an (outdated) introduction is found in doc/blockdolfin.pdf. Familiarity with the DOLFIN python interface is required. For more details of use, I recommend looking at the demos (start with demo/mixedpoisson.py), and the comments therein.

Bugs, questions, contributions: Visit http://github.com/blocknics/cbc.block.

The code is licensed under the GNU Lesser Public License, found in COPYING, version 2.1 or later. Some files under block/iterative/ use the BSD license, this is noted in the individual files.

Status (master branch)

Full test coverage report

Installation

The fenics-dolfin package is available from conda-forge, or preinstalled on Docker images from https://quay.io/organization/fenicsproject. The recipe below assumes that you have an existing conda-forge installation; to install on Docker images, skip the first few steps.

# Create an environment

conda create cbc-block
conda activate cbc-block

# Install base and dolfin (and optionally pytrilinos, but it doesn't seem to have a compatible version atm)

conda install pip fenics-dolfin scipy

# Install haznics from source. Examples are in examples/haznics.
# This can be skipped, but some demos will fail to run.

HAZ_VER=1.0.1
git clone --branch v${HAZ_VER} --depth 1 https://github.com/HAZmathTeam/hazmath
cd hazmath

conda install compilers c-compiler cxx-compiler fortran-compiler cmake>=3.15 make swig
sed -i -e '/cmake_minimum_required/s/3.12/3.15/' CMakeLists.txt
make config shared=yes suitesparse=yes lapack=yes haznics=yes swig=yes
make install
cp -a swig_files haznics
mv haznics/haznics.py haznics/__init__.py
cat >setup.py <<-EOF
	from distutils.core import setup
	setup(name='haznics', version='${HAZ_VER}', packages=['haznics'],
          package_data={'haznics': ['_haznics.so']})
EOF
python -m pip install .

# Install cbc.block itself. To install from source,
# use "git clone https://..." followed by "pip install -e <dir>[haznics]" instead.

pip install "cbc.block[haznics] @ git+https://github.com/blocknics/cbc.block"

Publications

1. K.-A. Mardal and J. B. Haga (2012). Block preconditioning of systems of PDEs. In A. Logg, K.-A. Mardal, G. N. Wells et al. (ed) Automated Solution of Differential Equations by the Finite Element Method, Springer. doi:10.1007/978-3-642-23099-8_35, http://fenicsproject.org/book
2. A. Budisa, X. Hu, M. Kuchta, K.-A. Mardal and L. Zikatanov (2022). HAZniCS — Software Components for Multiphysics Problems. doi: 10.48550/ARXIV.2210.13274, https://arxiv.org/abs/2210.13274