Implementation and Evaluation of GBDI Memory Compression Algorithm Using C/C++ on a Broader Range of Workloads

TL;DR

This paper implements and evaluates the GBDI memory compression algorithm in C/C++ across diverse workloads, demonstrating its effectiveness in reducing memory footprint and enhancing system performance.

Contribution

It provides a C/C++ implementation of GBDI and assesses its performance on a wider range of workloads, offering insights into its practical benefits.

Findings

GBDI achieves significant compression ratios across workloads

The C/C++ implementation allows for workload-specific customization

Memory compression with GBDI improves system performance

Abstract

Memory compression is an important approach in computer architecture for decreasing memory footprint and improving system performance. In this paper, we use C/C++ to develop a current memory compression algorithm; the Global Bases Delta Immediate (GBDI) algorithm, which was proposed at HPCA'2022. By using global bases and enabling deltas within the same block to vary in size, the GBDI compression algorithm decreases the size of encoded data. The goal of this research is to assess the effectivenessof the GBDI algorithm by examining its compression ratios under a broader range of workloads. Our research leads to a better knowledge of the GBDI algorithm's effectiveness and the potential benefits of memory compression techniques for various sorts of applications. Furthermore, our C/C++ version of the algorithm gives academics and practitioners a high degree of freedom over customizing the algorithm for individual workloads and optimizing its performance.