MetaDSE: A Few-shot Meta-learning Framework for Cross-workload CPU Design Space Exploration

2025-04-18

AI Summary

In CPU design space exploration (DSE), existing methods suffer from poor cross-workload generalization, high simulation overhead for target workloads, susceptibility to overfitting, and limited knowledge transferability. To address these challenges, this paper proposes the first few-shot meta-learning framework tailored for CPU DSE. Its key contributions are: (1) a novel workload-adaptive learnable architecture mask that explicitly models the coupling between hardware configurations and workload characteristics; and (2) the first integration of model-agnostic meta-learning (MAML) into CPU DSE, enabling rapid adaptation to unseen workloads with only 3–5 simulations. Evaluated on the SPEC CPU 2017 benchmark, our method reduces prediction error by 44.3% over state-of-the-art approaches, significantly improving both cross-workload generalization and optimization efficiency. The implementation is publicly available.

Technology Category

• Machine Learning: Hardware-aware ML
• Search and Optimization: Metareasoning and Metaheuristics
• Cognitive Modeling & Cognitive Systems: (Computational) Cognitive Architectures

Application Category

• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines

Abstract

Cross-workload design space exploration (DSE) is crucial in CPU architecture design. Existing DSE methods typically employ the transfer learning technique to leverage knowledge from source workloads, aiming to minimize the requirement of target workload simulation. However, these methods struggle with overfitting, data ambiguity, and workload dissimilarity. To address these challenges, we reframe the cross-workload CPU DSE task as a few-shot meta-learning problem and further introduce MetaDSE. By leveraging model agnostic meta-learning, MetaDSE swiftly adapts to new target workloads, greatly enhancing the efficiency of cross-workload CPU DSE. Additionally, MetaDSE introduces a novel knowledge transfer method called the workload-adaptive architectural mask algorithm, which uncovers the inherent properties of the architecture. Experiments on SPEC CPU 2017 demonstrate that MetaDSE significantly reduces prediction error by 44.3% compared to the state-of-the-art. MetaDSE is open-sourced and available at this href{https://anonymous.4open.science/r/Meta_DSE-02F8}{anonymous GitHub.}

Problem

• Addresses overfitting in cross-workload CPU design exploration
• Improves efficiency using few-shot meta-learning for new workloads
• Reduces prediction error with workload-adaptive architectural masks

Innovation

• Few-shot meta-learning for CPU DSE
• Workload-adaptive architectural mask algorithm
• Model agnostic meta-learning adaptation