Phases, Modalities, Temporal and Spatial Locality: Domain Specific ML Prefetcher for Accelerating Graph Analytics

TL;DR

This paper introduces MPGraph, an ML-based prefetcher tailored for graph analytics that employs domain-specific models and novel optimizations to significantly improve memory performance and IPC in graph processing tasks.

Contribution

It proposes MPGraph with three innovative optimizations, achieving higher prediction accuracy and IPC gains over existing prefetchers and ML models in graph analytics.

Findings

Transition detector outperforms Kolmogorov-Smirnov Windowing and decision tree.

Predictors achieve higher F1-score and accuracy-at-10 than LSTM and attention models.

MPGraph with CSTP yields 12.52-21.23% IPC improvement, surpassing state-of-the-art prefetchers.

Abstract

Memory performance is a bottleneck in graph analytics acceleration. Existing Machine Learning (ML) prefetchers struggle with phase transitions and irregular memory accesses in graph processing. We propose MPGraph, an ML-based Prefetcher for Graph analytics using domain specific models. MPGraph introduces three novel optimizations: soft detection for phase transitions, phase-specific multi-modality models for access delta and page predictions, and chain spatio-temporal prefetching (CSTP) for prefetch control. Our transition detector achieves 34.17-82.15% higher precision compared with Kolmogorov-Smirnov Windowing and decision tree. Our predictors achieve 6.80-16.02% higher F1-score for delta and 11.68-15.41% higher accuracy-at-10 for page prediction compared with LSTM and vanilla attention models. Using CSTP, MPGraph achieves 12.52-21.23% IPC improvement, outperforming state-of-the-art non-ML prefetcher BO by 7.58-12.03% and ML-based prefetchers Voyager and TransFetch by 3.27-4.58%. For practical implementation, we demonstrate MPGraph using compressed models with reduced latency shows significantly superior accuracy and coverage compared with BO, leading to 3.58% higher IPC improvement.