Distributed Simulation of Multicomputer Architectures with Mermaid

This paper describes the parallelization of the Mermaid multicomputer simulation environment. Due to our simulation methodology, this parallelization is reasonably straightforward as it does not require any measures to guarantee the causality within the simulated system. The resulting distributed simulator increases the simulation performance without any loss of simulation accuracy. Furthermore, the parallel simulation environment is also more scalable than its sequential counterpart with respect to the memory requirements. This gain in performance and scalability can be used for simulating larger target architectures and more realistic applications. Experiments with parallel Mermaid confirm that distributed simulation may lead to significant performance improvements compared to sequential simulation. In several cases, we even measured super-linear speedups.