2025.2.0 - Feb 2026
- Fixed handling of four- and five-point water models #2.
- Add support for the OpenCL platform and optimise GPU kernels #6.
- Clamp exponent to avoid exponential overflow when calculation the acceptance probability for the PME correction #9.
- Reduce memory footprint by using a shared primary CUDA context on each GPU device #10.
2025.1.0 - Nov 2025
- Initial public release.