Persistent and Partitioned MPI for Stencil Communication

TL;DR

This paper evaluates how persistent and partitioned MPI communication optimizations improve stencil operation performance, demonstrating up to 68% speedup in large-scale parallel applications.

Contribution

It provides a detailed analysis of the performance benefits of persistent and partitioned MPI communication routines for stencil operations at various scales.

Findings

Persistent MPI can speed up communication by up to 37%.

Partitioned MPI can speed up communication by up to 68%.

Performance improvements depend on process count, thread count, and message size.

Abstract

Many parallel applications rely on iterative stencil operations, whose performance are dominated by communication costs at large scales. Several MPI optimizations, such as persistent and partitioned communication, reduce overheads and improve communication efficiency through amortized setup costs and reduced synchronization of threaded sends. This paper presents the performance of stencil communication in the Comb benchmarking suite when using non blocking, persistent, and partitioned communication routines. The impact of each optimization is analyzed at various scales. Further, the paper presents an analysis of the impact of process count, thread count, and message size on partitioned communication routines. Measured timings show that persistent MPI communication can provide a speedup of up to 37% over the baseline MPI communication, and partitioned MPI communication can provide a speedup of up to 68%.