Techniques to enhance performance of an existing aviation simulation

Facing a need to run large scenarios on aviation models more quickly than the one to two days currently required, the MITRE Corporation undertook an effort to reduce the execution time of one such simulation. Time and cost constraints prohibited a major rewrite of the almost one million existing lines of code, so only solutions requiring minimal changes to the code base were considered. This paper describes the approaches taken to increase the speed of the original sequential simulation by employing more efficient algorithms and parallel processing technology. Specifically, an implementation of a new technique for parallel proximity detection provided an 80% reduction in the time spent checking for conflicts. In addition, implementation of a thread pool that enables the movement of multiple aircraft in parallel resulted in a 10%-15% reduction in the overall execution time of the simulation. In this paper we report on the design of these techniques and how they were implemented in the simulation.