<!-- https://docs.dseams.info/user/howto/engine-openmm.md -->

# OpenMM {#openmm}

Build the reporter against conda-forge OpenMM 8.6.1 or a later 8
release. Eigen and a C++ compiler need to be on the path. Configure
meson from that shell so `python3` imports `openmm`, and the plugin
looks for `libOpenMM` on the same prefix. The exports are the include
and library paths, including the AVX directories conda-forge uses.

```bash
pixi init -c conda-forge dseams-openmm
pixi add --manifest-path dseams-openmm/pixi.toml "openmm>=8.6.1,<9"
eval "$(pixi shell-hook --manifest-path dseams-openmm/pixi.toml)"
export PKG_CONFIG_PATH="$CONDA_PREFIX/lib/pkgconfig:$CONDA_PREFIX/lib.AVX2_256/pkgconfig:$CONDA_PREFIX/lib.AVX_256/pkgconfig${PKG_CONFIG_PATH:+:$PKG_CONFIG_PATH}"
export LIBRARY_PATH="$CONDA_PREFIX/lib:$CONDA_PREFIX/lib.AVX2_256${LIBRARY_PATH:+:$LIBRARY_PATH}"
export CPATH="$CONDA_PREFIX/include${CPATH:+:$CPATH}"
meson setup build -Dwith_openmm=enabled
meson compile -C build
```

`DseamsReporter` reads the OpenMM context in angstroms and calls
`InSitu.observe`. Loading `input/traj/ice1h.lammpstrj` onto
a Reference-platform context with a zero force and stepping once prints:

```text
openmm hexagonal 128 six-rings 256 q6 0.625855
```

The q6 mean is `seams_frame_steinhardt_graph` on the positions the
reporter classified, cutoff graph, degree 6.
