Collective Asynchronous Remote Invocation (CARI) Schedules

Abstract

Gadget-2 is a production parallel code for cosmological N-body and hydrodynamic simulations. Versions of this code featured in the Millennium Simulation, which is the largest simulated model of the universe. Gadget-2 has been parallelized for distributed memory platforms using the MPI standard. In this thesis | we analyzed Gadget-2 with a view to understanding what high-level SPMD communication abstractions might be developed to replace the intricate use of MPI in such irregular applications and do so without compromising the e ciency. Our analysis revealed that the use of low-level MPI primitives bundled with the computation code makes Gadget-2 di cult to understand and probably hard to maintain. In addition, we found out that the original Gadget-2 code contains a small handful of complex and recurring-patterns of message passing. We also noted that these complex patterns can be reorganized into a higher level communication library with some modi cations to the Gadget-2 code. This thesis presents the implementation and evaluation of one such message passing pattern (or schedule) that we term Collective Asynchronous Remote Invocation (CARI). As the name suggests, CARI is a collective variant of Remote Method Invocation (RMI), which is an attractive, high-level, and established paradigm in distributed systems programming. The CARI API might be implemented in several ways | we implement and evaluate two versions of this API on a compute cluster. The performance evaluation reveals that CARI versions of the Gadget-2 code perform as good as the original Gadget-2 code but the level of abstraction is raised considerably. In fact one of the implementations has geared towards scalability on larger number of cores which performs the best when the problem size reaches its scalability limits.