Phases, Modalities, Temporal and Spatial Locality: Domain Specific ML Prefetcher for Accelerating Graph Analytics
This addresses memory performance issues for graph processing systems, offering a domain-specific solution with incremental improvements over existing ML methods.
The paper tackles memory bottlenecks in graph analytics by proposing MPGraph, a domain-specific ML prefetcher that introduces optimizations like soft phase detection and chain spatio-temporal prefetching, achieving up to 21.23% IPC improvement and outperforming state-of-the-art prefetchers by up to 12.03%.
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.