一句话：提出配置墙量化模型与 MLIR 编译优化，消除或隐藏加速器配置开销，在 OpenGeMM 上几何平均加速 2 倍。
现代计算平台将内核卸载到集成加速器以提升能效，但 CPU 用于初始化、控制和同步的时间随加速器复杂度增长，导致配置受限，即“配置墙”。本文首先提出一种广泛适用的 roofline 模型变体，用于量化系统何时受配置约束；随后提出面向领域的编译器抽象及配套优化 pass，并在 MLIR 中实现，以在开源架构上生成优化二进制。实验在 OpenGeMM 上通过消除冗余配置周期并自动隐藏剩余配置周期，取得几何平均 2 倍性能提升。工作为理解加速器配置机制对性能的影响及自动规避配置墙提供了关键洞见。
Contemporary compute platforms increasingly offload compute kernels from CPU to integrated hardware accelerators to reach maximum performance per Watt. Unfortunately, the time the CPU spends on setup control and synchronization has increased with growing accelerator complexity. For systems with complex accelerators, this means that performance can be configuration-bound. Faster accelerators are more severely impacted by this overlooked performance drop, which we call the configuration wall. Prior work evidences this wall and proposes ad-hoc solutions to reduce configuration overhead. However, these solutions are not universally applicable, nor do they offer comprehensive insights into the underlying causes of performance degradation. In this work, we first introduce a widely-applicable variant of the well-known roofline model to quantify when system performance is configuration-bound. To move systems out of the performance-bound region, we subsequently propose a domain-specific compiler abstraction and associated optimization passes. We implement the abstraction and passes in the MLIR compiler framework to run optimized binaries on open-source architectures to prove its effectiveness and generality. Experiments demonstrate a geomean performance boost of 2x on the open-source OpenGeMM system, by eliminating redundant configuration cycles and by automatically hiding the remaining configuration cycles. Our work provides key insights in how accelerator performance is affected by setup mechanisms, thereby facilitating automatic code generation for circumventing the configuration wall.
DOI 原文 · arXiv · PDF（开放获取） · DBLP
@inproceedings{DelmLDCYAWGV26,
title = {The Configuration Wall: Characterization and Elimination of Accelerator Configuration Overhead},
author = {Josse Van Delm and Anton Lydike and Joren Dumoulin and Jonas Crols and Xiaoling Yi and Ryan Antonio and Jackson Woodruff and Tobias Grosser and Marian Verhelst},
booktitle = {ASPLOS},
year = {2025},
doi = {10.1145/3760250.3762225}
}