Composing Distributed Computations Through Task and Kernel Fusion

TL;DR

Diffuse is a system that enables dynamic task and kernel fusion in distributed runtime environments, significantly improving performance and scalability of task-based applications across multiple GPUs.

Contribution

Diffuse introduces a scalable intermediate representation for distributed computation that facilitates task and kernel fusion, enhancing performance across real-world libraries.

Findings

Achieves 1.86x average speedup on unmodified applications

Enables performance scaling up to 128 GPUs

Identifies optimization opportunities missed by developers

Abstract

We introduce Diffuse, a system that dynamically performs task and kernel fusion in distributed, task-based runtime systems. The key component of Diffuse is an intermediate representation of distributed computation that enables the necessary analyses for the fusion of distributed tasks to be performed in a scalable manner. We pair task fusion with a JIT compiler to fuse together the kernels within fused tasks. We show empirically that Diffuse's intermediate representation is general enough to be a target for two real-world, task-based libraries (cuNumeric and Legate Sparse), letting Diffuse find optimization opportunities across function and library boundaries. Diffuse accelerates unmodified applications developed by composing task-based libraries by 1.86x on average (geo-mean), and by between 0.93x--10.7x on up to 128 GPUs. Diffuse also finds optimization opportunities missed by the original application developers, enabling high-level Python programs to match or exceed the performance of an explicitly parallel MPI library.