Merging parallel simulation programs

In earlier work cloning is proposed as a means for efficiently splitting a running simulation midway through its execution into multiple parallel simulations. In simulation cloning, clones usually are able to share computations that occur early in the simulation, but as their states diverge individual LPs are replicated as necessary so that their computations proceed independently. However, if, over time the state of the clones (or their constituent LPs) converges there is, as of yet, no means for recombining them. In this case some efficiency is lost because they will execute identical events. This idea is the reverse of cloning, as we merge logical processes that have been previously cloned and we show that this can further increase efficiency because the new un-cloned LPs will complete computations that would otherwise be duplicated. We discuss our implementation of merging, and illustrate its effectiveness in several example simulation scenarios.