Cache-aware load-balancing mechanisms for synchronous computations on shared-memory multiprocessors

Synchronous algorithms constitute an important class of algorithms for parallel and distributed processing. We present several cache-aware load-balancing mechanisms for synchronous computations on shared-memory multiprocessors. We show both theoretically and empirically that the mechanisms incur a very low overhead. The mechanism is tested with a number of applications that rely on synchronous computations. The results confirm that one of the mechanisms proposed is highly efficient and is scalable to more processors.