Improving compiler and run-time support for irregular reductions using local writes

Abstract

Current compilers for distributed-memory multiprocessors parallelize irregular reductions either by generating calls to sophisticated run-time systems (CHAOS) or by relying on replicated buffers and the shared-memory interface supported by software DSMs (TreadMarks). We introduce LocalWrite, a new technique for parallelizing irregular reductions based on the owner-computes rule. It eliminates the need for buffers or synchronized writes, but may replicate computation. We investigate the impact of connectivity (node/edge ratio), locality (accesses to local data) and adaptivity (edge modifications) on their relative performance. LocalWrite improves performance by 50{150% compared to using replicated buffers, and can match or exceed gather/scatter for applications with low locality or high adaptivity.