A Server-based Approach for Predictable GPU Access with Improved Analysis

TL;DR

This paper introduces a server-based GPU management approach that ensures predictable access, reduces busy waiting and priority inversion, and improves task schedulability in real-time systems, demonstrated through a prototype on an embedded platform.

Contribution

The paper presents a novel server-based GPU management method that improves predictability and efficiency over traditional synchronization approaches in real-time systems.

Findings

Significant improvement in task schedulability with the server-based approach

Reduction in busy waiting and priority inversion issues

Practical implementation demonstrating effectiveness on embedded hardware

Abstract

We propose a server-based approach to manage a general-purpose graphics processing unit (GPU) in a predictable and efficient manner. Our proposed approach introduces a GPU server that is a dedicated task to handle GPU requests from other tasks on their behalf. The GPU server ensures bounded time to access the GPU, and allows other tasks to suspend during their GPU computation to save CPU cycles. By doing so, we address the two major limitations of the existing real-time synchronization-based GPU management approach: busy waiting and long priority inversion. We have implemented a prototype of the server-based approach on a real embedded platform. This case study demonstrates the practicality and effectiveness of the server-based approach. Experimental results indicate that the server-based approach yields significant improvements in task schedulability over the existing synchronization-based approach in most practical settings. Although we focus on a GPU in this paper, the server-based approach can also be used for other types of computational accelerators.