Installing binaries¶
MPI parallel builds¶
You must build HOOMD-blue from source to enable support for the native MPI and CUDA libraries on your HPC resource. You can use the glotzerlab-software repository to manage such builds as conda packages.
Serial CPU and single GPU builds¶
HOOMD-blue binaries for serial CPU and single GPU are available on conda-forge for the
linux-64, osx-64, and osx-arm64 platforms. Install the hoomd
package from the conda-forge
channel into a conda environment:
$ mamba install hoomd=4.7.0
conda
auto-detects whether your system has a GPU and attempts to install the appropriate
package. Override this and force the GPU enabled package installation with:
$ export CONDA_OVERRIDE_CUDA="12.0"
$ mamba install "hoomd=4.7.0=*gpu*" "cuda-version=12.0"
Similarly, you can force CPU only package installation with:
$ mamba install "hoomd=4.7.0=*cpu*"
Note
CUDA 11.8 compatible packages are also available. Replace “12.0” with “11.8” above when installing HOOMD-blue on systems with CUDA 11 compatible drivers.
Note
Run time compilation is no longer available on conda-forge builds starting with HOOMD-blue 4.7.0.
Tip
Use miniforge, miniconda, or any other minimal conda environment provider instead of the full Anaconda distribution to avoid package conflicts with conda-forge packages. When using miniconda, follow the instructions provided in the conda-forge documentation to configure the channel selection so that all packages are installed from the conda-forge channel.