Optimizing Iterative Data-flow Scientific Applications using Directed Cyclic Graphs

Abstract

Listen

Translate article

Data-flow programming models have become a popular choice for writing parallel applications as an alternative to traditional work-sharing parallelism. They are better suited to write applications with irregular parallelism that can present load imbalance. However, these programming models suffer from overheads related to task creation, scheduling and dependency management, limiting performance and scalability when tasks become too small. At the same time, many HPC applications implement iterative methods or multi-step simulations that create the same directed acyclic graphs of tasks on each iteration. By giving application programmers a way to express that a specific loop is creating the same task pattern on each iteration, we can create a single task DAG once and transform it into a cyclic graph. This cyclic graph is then reused for successive iterations, minimizing task creation and dependency management overhead. This paper presents the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">taskiter</i> , a new construct we propose for the OmpSs-2 and OpenMP programming models, allowing the use of directed cyclic task graphs (DCTG) to minimize runtime overheads. Moreover, we present a simple <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">immediate successor</i> locality-aware heuristic that minimizes task scheduling overhead by bypassing the runtime task scheduler. We evaluate the implementation of the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">taskiter</i> and the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">immediate successor</i> heuristic in 8 iterative benchmarks. Using small task granularities, we obtain a geometric mean speedup of 2.56x over the reference OmpSs-2 implementation, and a 3.77x and 5.2x speedup over the LLVM and GCC OpenMP runtimes, respectively.