Documentation Index
Fetch the complete documentation index at: https://mintlify.com/DedalusProject/dedalus_conda/llms.txt
Use this file to discover all available pages before exploring further.
Overview
The Dedalus build scripts provide flexible MPI (Message Passing Interface) configuration options. You can either install OpenMPI from conda-forge or link to a custom system MPI installation.Configuration Options
INSTALL_MPI
Controls whether to install OpenMPI from conda-forge.Valid values:
1- Install OpenMPI from conda-forge (default)0- Use custom MPI installation (requires MPI_PATH)
MPI_PATH
Path to your custom MPI installation prefix.Must be set when
INSTALL_MPI=0. The build script will look for MPI binaries in ${MPI_PATH}/bin.Installation Modes
- OpenMPI from conda
- Custom MPI
The default configuration installs OpenMPI from conda-forge along with the required compilers and mpi4py.This mode:
- Installs conda-forge compilers
- Installs OpenMPI and openmpi-mpicc
- Installs mpi4py from conda
- Works seamlessly with conda-installed FFTW and HDF5
Build Behavior
With INSTALL_MPI=1
From the build script:conda_install_dedalus3.sh
compilers- Conda-forge compiler toolchainopenmpi- OpenMPI libraryopenmpi-mpicc- MPI C compiler wrappermpi4py- Python MPI bindings
With INSTALL_MPI=0
From the build script:conda_install_dedalus3.sh
- Adds your custom MPI binaries to PATH
- Verifies
mpiccis accessible - Builds mpi4py from source against your custom MPI
Validation
The build script validates MPI configuration before proceeding:conda_install_dedalus3.sh
Platform Considerations
HPC Clusters
On HPC systems, you typically want to use the system-provided MPI:Workstations and Laptops
For local development, the conda-forge OpenMPI is recommended:Compatibility with FFTW
FFFTW from conda-forge is built for OpenMPI. If you use a custom MPI that is not OpenMPI-compatible, you may need to also use a custom FFTW build.From the build script comments:
Related Configuration
- FFTW Configuration - FFTW depends on MPI configuration
- HDF5 Configuration - HDF5 parallel support requires MPI