Using the Docker Image¶
Using the official runtime Docker image allows FrontISTR to be run without installing its dependency libraries. For installation of Docker itself, refer to the documentation for your environment.
Obtaining the Image¶
The image includes the following libraries required for execution.
- MPI: Open MPI
- METIS
- MUMPS
- Trilinos ML
- Intel MKL (x86_64 environments)
Running an Analysis¶
Run the analysis by sharing the host working directory (where the input files are located) with the container. Use -v to share the directory and -w to specify the working directory.
docker run --rm -v $PWD:/work -w /work \
registry.gitlab.com/frontistr-commons/frontistr/fistr1:latest fistr1
The default command for the image is fistr1 (source: docker/container/Dockerfile, CMD ["fistr1"]). To provide options such as a version check, specify the command after the image name.
For preparing input files and writing control files, see Analysis Flow.
MPI Parallel Execution¶
MPI parallel execution is performed through mpirun inside the container. For specific execution commands and the domain-decomposition procedure, see Parallel Analysis.