mdcore | |
About Features Download Publications Documentation Tutorial |
The following is a list of some of the most important features of mdcore, in no particular order: Arbitrary potential functions: The inter-particle potentials are evaluated using smooth piecewise polynomial interpolants which can be constructed from any user-supplied interaction function. Pairwise Verlet lists: Per-cell pair Verlet lists can be used to speed up low-density simulations without affecting parallel performance. POSIX pthreads: All multi-threading is implemented using the standard pthreads routines which should be available on any POSIX-compliant system. Distributed-memory parallelism: Basic distributed-memory parallelism is implemented with asynchronous MPI routines. Support for the Cell/BE: When configured with the "--with-cell" option, the library can compute non-bonded interactions on the SPUs as well as on the PPU. Support for CUDA GPUs: When configured with the "--with-cuda" option, the library can compute non-bonded interactions on a CUDA-capable GPU. Support for SSE/SSE2: When available, the interaction routines are compiled using SSE (single-precision) or SSE2 (double-precision) vectorized instructions. |