Problem with USER-CUDA EAM

I'm not sure if this has been fixed or not, but seemed like a good place to advertise that Trung Dac Nguyen (ORNL) has implemented an alternative EAM style for the GPU package ( eam/gpu, eam/fs/gpu, and eam/alloy/gpu ). This version can perform better depending on the system/setup and has several advantages:

- Support for hybrid pair styles
- Support for minimization in LAMMPS
- Deterministic results
- Mixed precision calculations (single precision w/ double precision accumulation)
- Support for OpenCL in addition to CUDA

It can be downloaded here (under Current Release):

and I will also be sending a patch to Steve. - Mike