TaskTorrent: a Lightweight Distributed Task-Based Runtime System in C++

TL;DR

TaskTorrent is a lightweight, distributed C++14 runtime system that efficiently manages task graphs using MPI, enabling scalable execution of large linear algebra problems on thousands of cores.

Contribution

It introduces a novel distributed task-based runtime with a fully distributed task graph discovery mechanism in C++14 using MPI.

Findings

Minimal overhead compared to existing solutions

Excellent scalability to thousands of cores

Effective application to large linear algebra problems

Abstract

We present TaskTorrent, a lightweight distributed task-based runtime in C++. TaskTorrent uses a parametrized task graph to express the task DAG, and one-sided active messages to trigger remote tasks asynchronously. As a result the task DAG is completely distributed and discovered in parallel. It is a C++14 library and only depends on MPI. We explain the API and the implementation. We perform a series of benchmarks against StarPU and ScaLAPACK. Micro benchmarks show it has a minimal overhead compared to other solutions. We then apply it to two large linear algebra problems. TaskTorrent scales very well to thousands of cores, exhibiting good weak and strong scalings.