This molecular dynamics simulation program is accelerated by using the power of general purpose graphic processing units (GPGPUs). Due to the highly parallel nature of such graphic cards and the availability of a fairly large amount of shared memory, molecular dynamic simulations can harvest these features and gain a huge speedup.
For the current simulation, the C to CUDA interface delivered by NVIDIA is used.
The workflow of the program is as follows: First, the simulation parameter given at the command line are evaluated and the initial configuration is read in. Then, all relevant data is transfered form the host memory to the (graphic card) device memory. As from now, the whole MD simulation is done on the GPU and only intermediate results are transfered back to the host and saved into output files.
In the current setup, the code is limited to simulate the elastic Lennard-Jones model as descibed by Peleg et. al., but it can be easily adapted to other models.
It is recommended, to follow the next three links in order to lern how to use this program, it's license and credits:
1.6.1
