Toward petascale cosmological simulations with changa

Cosmological simulators are currently an important component in the study of the formation of galaxies and large scale structures, and can help answer many important questions about the universe. Despite their utility, existing parallel simulators do not scale effectively on modern machines containing thousands of processors. In this paper we present ChaNGa, a recently released production simulator based on the Charm++ infrastructure. To achieve scalable performance, ChaNGa employs various optimizations that maximize the overlap between computation and communication. Other optimizations described in the paper include a particle caching mechanism and a recently added multi-timestepping scheme. The paper shows experimental results demonstrating scalable performance for simulations on machines with thousands of processors, including IBM-BlueGene/L and Cray-XT3. We conclude by pointing to our current work, which focuses on developing new load balancing strategies to further improve ChaNGa’s parallel efficiency. To our knowledge, ChaNGa is presently the most scalable cosmological simulator available.