arXiv cs.AI· Srinidhi N, Sunanda Somwase, Hemant Kumar Chandrakant Chandavar, Eashan Dash, Varun Agrawal, Prakash Raghavendra·· 14 小时前AI 评分31
Waypoint：为 CPU 上的 GEMM 建立解析模型
Waypoint: Analytically Modeling GEMM for CPUs
AI 导读
Waypoint 是一个解析模型，可为给定硬件和输入矩阵形状自动搜索最优 GEMM 分块尺寸，估算时涵盖存储层次交互、指令级并行与多线程等架构效应。基于该模型构建的 CPU GEMM 库 OPAM 性能可匹敌当前最优 GEMM 库，并将 CPU 上的 LLM 推理加速 1.36-1.98 倍。
正文 · 原文
Abstract:High-performance General Matrix Multiplication (GEMM) implementations rely on breaking down GEMM into carefully selected tiles to achieve peak performance. We present Waypoint, a new analytical model to automatically identify the optimal tile size for a given hardware and input matrix shape. Waypoint estimates the achievable performance for a given tile choice by capturing all the architectural effects including interactions with memory hierarchy, instruction-level parallelism, and multi-threading. Using the estimates as a proxy, we search through the space of possible tile sizes to identify the one that maximizes performance. Additionally, Waypoint helps hardware designers estimate realistic performance on experimental hardware optimizations. We build a new GEMM library OPAM for CPUs that uses Waypoint to automatically break down GEMM and achieve performance that matches state-of-the-art GEMM libraries and helps accelerate LLM inference on CPUs achieving $1.36 - 1.98 \times$ speedups.
| Subjects: | Distributed, Parallel, and Cluster Computing (cs.DC); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.34914 [cs.DC] |
| (or arXiv:2610.34914v1 [cs.DC] for this version) | |
| https://doi.org/10.48550/arXiv.2610.34914 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Srinidhi Nagendra [view email]
[v1]
Mon, 26 Oct 2026 06:40:51 UTC (265 KB)
来源：arXiv cs.AI · arxiv.org