Graphics Processing Unit (GPU) Acceleration for Cosite Interference Prediction Tools

Simulation tools that exist today for prediction of cosite interference in complex RF environments are able to accurately model complex platforms but they are quickly constrained by computational limits on traditional workstations. The focus of this research will be on updating existing electromagnetic interference and vulnerability (EMI/EMV) simulation tools to the latest parallel computing clusters. As a result of this effort the time required for analyzing cosite interference on complex platforms will be significantly reduced. ANDRO will build upon its existing E3Expert framework and integrate a new EM analysis tool, uCAST (UTD Complex Antenna Simulation Tool). uCAST is a UTD (Uniform Theory of Diffraction) based code developed by Applied EM, Inc. that is currently being enhanced with parallel processing capabilities under a related NAVAIR SBIR Phase II activity. As part of that effort bottlenecks in current algorithms were identified as well as other algorithms which may be problematic in transferring to a parallel environment. That work proceeded into Phase II where a commercial grade GPU based code was developed using NVIDIA"s CUDA platform. In this Phase I effort we will assess the feasibility of integrating the uCAST kernel into the E3Expert and develop a work plan for building requirements to test and prototype its interface. During Phase II we plan to develop a commercialized version of E3Expert built with uCAST"s UTD based parallel processing capabilities for cosite analysis.