PyPIC Version 2.2.1
PyPIClib GPU module hotfixes:
- only define double precision atomicAdd for CUDA <v8.0 (used in particle-to-mesh functions, FP64 atomicAdd has been included in CUDA v8.0 and is hardware accelerated from computing capability 6.0 on, e.g. for the NVIDIA P100)
- [slow] CPU versions of particle-to-mesh and mesh-to-particle functions now check whether the particle is on the mesh (usual behaviour for GPU)