Cell automaton modelling algorithms: Implementation and testing in GPU systems

The architecture of today's video cards is able to execute up to hundreds of thousands of operation in parallel. This ability creates the possibility to solve computationally intensive tasks with minimal effort. Our research aims to investigate how to use the graphics hardware for general computing ability in biological models. In the development we have used a re-thought, and upgraded successor of the Nvidia G80 architecture, Fermi-GF104 architecture, and the associated CUDA programming environment in C/C++ language environment. After the developer machine and the test environment were complied, a general cellular automaton modelling framework was developed. It is solved partly by parallel algorithm because it calculates on matrix data structure. Several parallel algorithms and data were tested using the system. The speed of program execution was measured and the CGMA (compute to global memory access) ratio was determined. Compared to the performance of the serial execution we experienced an order of magnitude increase.