Dynamic translation for virtual machine based traffic simulation

For virtual machine based traffic simulation platforms, the paper proposes a software framework that performs trace-based dynamic translation. Through monitoring the runtime execution status of bytecodes and translating frequently executed bytecodes, also known as hot spots, into equivalent native machine codes, the framework considerably improves the performance of virtual machine based traffic simulation platforms up to ten times or more, as the experiments showed. For the first time, the presented work clearly exhibits that a seamless combination of the two technologies – dynamic translation and virtual machine could lead to a new generation of applicable traffic simulation platforms. Such a platform not only offers high flexibility in terms of traffic model simulation, but also preserves the ability of conducting numerical computation-intensive simulations generally found in real-life industrial projects.