A Massively Parallel Implementation of Gillespie Algorithm on FPGAs

This paper targets the acceleration of complex stochastic simulations of biochemical systems by a dedicated hardware architecture on configurable devices (FPGA). Existing approaches are discussed and compared with the proposed one, and experimental data is introduced to support the feasibility of the system. Retargetable hardware description can be automatically generated for any suitable simulation problem, and preliminary results show very high performance - 100 million time steps per second for large models (1000 reactions).