Skip to content

Latest commit

 

History

History
214 lines (144 loc) · 7.36 KB

README.md

File metadata and controls

214 lines (144 loc) · 7.36 KB
Raw

This is a fork of https://bitbucket.org/pefarrell/defcon for the purposes of the ExCALIBUR Project NEPTUNE.

Synopsis

The defcon library implements the deflated continuation algorithm of Farrell, Beentjes and Birkisson. Its objective is to compute the solutions of

f(u, λ) = 0,

where u is the solution of a PDE and λ is a parameter on which the PDE depends.

The algorithm that defcon implements has two main advantages over the previous state of the art:

  • Defcon can compute disconnected bifurcation diagrams as well as connected ones.
  • The algorithm can scale to massive discretisations of PDEs if a scalable preconditioner is available.

For a full description of the algorithm, see

http://arxiv.org/abs/1603.00809

Dependencies

Defcon can be used either with Firedrake (http://firedrakeproject.org) or FEniCS (http://fenicsproject.org). FEniCS-X is not yet supported. Development now mainly occurs with Firedrake.

If you're compiling FEniCS yourself, you'll need

Defcon recommends (and some of the examples depend on)

Current status and automated testing

Defcon is automatically tested against a development version of Firedrake, using Bitbucket Pipelines and CircleCI. This ensures that defcon should always run with a recent Firedrake development version.

Pipelines CircleCI

To run the tests yourself, do

mpiexec -n 2 py.test -x -v examples/firedrake

when everything is installed.

Code Examples

The easiest way to learn how to use it is to examine the examples in examples/. Start with examples/firedrake/elastica or examples/fenics/elastica, and compare to the Euler elastica section of the manuscript cited above.

Installation

pip install -e .

Use in anaconda environments with FEniCS

Here is an example of how to use defcon with anaconda and FEniCS:

# Install FEniCS
conda create -n fenicsproject -c conda-forge fenics
source activate fenicsproject

# Install other dependencies
conda install scipy
pip install h5py # conda-supplied h5py is incompatible with FEniCS
pip install https://bitbucket.org/mpi4py/mpi4py/get/master.tar.gz
conda install matplotlib scipy ipython pyqt

# Install defcon
cd /path/to/defcon
pip install -e .

# Try it out
cd examples/fenics/wingedcusp
mpiexec -n 2 python wingedcusp.py
defcon gui

The official defcon Docker container

A Docker container for defcon is generated weekly as part of the CI pipeline. This container comes with the Firedrake backend.

The defcon Docker image can be run locally on a machine with Docker installed by executing:

docker run -it jdbetteridge/defcon

Each month a release is created and a specific release can be selected by adding the tag to the image name eg: jdbetteridge/defcon:2023-04 will use defcon from April 2023.

Inside the image defcon is installed in the /home/firedrake/firedrake/src/defcon directory. To run an example first activate the Firedrake virtual environment by executing:

source /home/firedrake/firedrake/bin/activate

The Bratu example can be run in the activated environment by executing:

cd $VIRTUAL_ENV/src/defcon/examples/firedrake/bratu
mpiexec -n 2 python bratu.py

Use in FEniCS Docker containers

FEniCS Docker containers introduce a convenient way of distributing FEniCS on many platforms, see

http://fenics-containers.readthedocs.io/en/latest/index.html .

To use defcon in a FEniCS docker container, simply fire up a container with development version of FEniCS, e.g. using fenicsproject run dev, and in the container type

CC=mpicc HDF5_MPI=ON pip3 install --no-binary=h5py --no-cache-dir --user h5py
pip3 install --user https://bitbucket.org/pefarrell/defcon/get/master.tar.gz

Then you can navigate to defcon demos and run them

cd /home/fenics/.local/share/defcon/examples/fenics/elastica
mpirun -n 2 python3 elastica.py

Running the GUI in docker

To use the defcon GUI, a much more complicated procedure is needed. We suggest installing FEniCS via anaconda as described above, instead. However, if you're very keen on GUIs + docker, here are some notes.

Connecting X11 applications on Linux

One needs to allow a docker container to connect to host's X11 system:

xhost +
docker run -ti -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix quay.io/fenicsproject/dev

or another docker image in place of quay.io/fenicsproject/dev. Don't forget to run xhost - when finished with the container.

Running the GUI in docker on OSX

If running on OSX, you need to have the appropriate XQuartz program running, and to convince it to talk to the GUI in docker. To do so, perform the following steps:

  • Run XQuartz

  • Go to Preferences > 'Security' tab; turn on 'Allow connection from network clients'

  • Restart XQuartz

  • Get your local machine's IP:

      ip=$(ifconfig en0 | grep inet | awk '$1=="inet" {print $2}') && echo "My IP is: $ip"

  • Allow the local machine to talk to XQuartz and run the docker container:

      xhost + ${ip}
  docker run -ti -e DISPLAY=${ip}:0 -v /tmp/.X11-unix:/tmp/.X11-unix quay.io/fenicsproject/dev /bin/bash

    or another docker image in place of quay.io/fenicsproject/dev.

Don't forget to run xhost - when finished with the container.

Running the GUI in the FEniCS-supplied images

In the container one needs to install latex and one of PyQt5, PyQt4, or PySide. The most convenient is

sudo apt update
sudo apt install -y python3-pyqt5 dvipng texlive-latex-extra texlive-fonts-recommended

After installing h5py and defcon as described above, one can run the gui and start a defcon application by

cd /home/fenics/.local/share/defcon/examples/fenics/elastica
export QT_GRAPHICSSYSTEM=native # may not be necessary on all systems
defcon gui &
mpirun -n 2 python3 elastica.py

Note that only aspects of FEniCS docker containers directly related to defcon have been shown. To setup a practical workflow (allowing preservation of JIT cache, etc.), we recommend the FEniCS Docker manual, at

http://fenics-containers.readthedocs.io/en/latest/index.html .

Troubleshooting

  • Make sure all mesh constructors, Expressions and CompiledSubDomains take in the MPI communicator (see e.g. examples/fenics/navier-stokes). This is the most common cause of silent deadlocks.

Contributors

License

GNU LGPL, version 3.