Hardware Implementation of Parallel SOARS using FPGA based Multiprocessor Architecture

Abstract

We are developing an agent-based simulation language, SOARS (Spot Oriented Agent Role Simulator), aiming standard descriptive language to express complicated social system. SOARS is a simulation language with original model concept to describe agent’s interaction. Currently, SOARS processor is implemented by JAVA, and agent rule is successively processed. However, model concept of SOARS is originally headed for parallel execution of agent rule, and it is preferable to prepare for high-speed implementation performing right parallel operation by common simulation description. Therefore, we decided to investigate the problem in language specification at the time of parallelizing, construct and monitor the system that is small-scaled but enables actual parallel operation.

Parallel processing with high granularity like the agent rule of SOARS is an interesting theme of computer science in itself. Such parallel processing is considered suitable to be implemented by multiprocessor core architecture CPU that is thought to come out one after another. The system available, however, is hard to obtain at present. Therefore, we developed unique processor architecture using FPGA (Field Programmable Gate Array) that recently shows remarkable progress and can be obtained at low cost.