DAG-based Scheduling with Resource Sharing for Multi-task Applications in a Polyglot GPU Runtime

TL;DR

This paper introduces a novel GPU runtime scheduler that enables efficient multi-task execution with resource sharing and overlap, improving performance and ease of use across multiple programming languages.

Contribution

A new GPU scheduler that provides transparent asynchronous execution and resource sharing without prior knowledge of program dependencies, integrated via the GrCUDA polyglot API.

Findings

44% average speedup over synchronous execution

No slowdown compared to hand-optimized CUDA Graphs code

Validated on 6 custom benchmarks for task-parallelism

Abstract

GPUs are readily available in cloud computing and personal devices, but their use for data processing acceleration has been slowed down by their limited integration with common programming languages such as Python or Java. Moreover, using GPUs to their full capabilities requires expert knowledge of asynchronous programming. In this work, we present a novel GPU run time scheduler for multi-task GPU computations that transparently provides asynchronous execution, space-sharing, and transfer-computation overlap without requiring in advance any information about the program dependency structure. We leverage the GrCUDA polyglot API to integrate our scheduler with multiple high-level languages and provide a platform for fast prototyping and easy GPU acceleration. We validate our work on 6 benchmarks created to evaluate task-parallelism and show an average of 44% speedup against synchronous execution, with no execution time slowdown compared to hand-optimized host code written using the C++ CUDA Graphs API.