SimNet: Accurate and High-Performance Computer Architecture Simulation using Deep Learning

TL;DR

SimNet introduces a GPU-accelerated, machine learning-based simulation framework that significantly speeds up computer architecture simulation while maintaining high accuracy, addressing the long runtime issue of traditional discrete-event simulators.

Contribution

The paper presents a novel ML-based instruction latency predictor integrated into a GPU-accelerated simulator, achieving faster and more accurate architecture simulations.

Findings

ML-based predictor improves simulation speed

GPU acceleration enhances throughput

High accuracy maintained in simulation results

Abstract

While discrete-event simulators are essential tools for architecture research, design, and development, their practicality is limited by an extremely long time-to-solution for realistic applications under investigation. This work describes a concerted effort, where machine learning (ML) is used to accelerate discrete-event simulation. First, an ML-based instruction latency prediction framework that accounts for both static instruction properties and dynamic processor states is constructed. Then, a GPU-accelerated parallel simulator is implemented based on the proposed instruction latency predictor, and its simulation accuracy and throughput are validated and evaluated against a state-of-the-art simulator. Leveraging modern GPUs, the ML-based simulator outperforms traditional simulators significantly.