RedFuser: An Automatic Operator Fusion Framework for Cascaded Reductions on AI Accelerators

TL;DR

RedFuser is a framework that automatically fuses cascaded reduction operations in AI models, significantly improving execution efficiency and surpassing existing compiler performance.

Contribution

It introduces a formal analysis methodology and an automated framework for general fusion of cascaded reductions in AI compilers.

Findings

Achieves 2x to 5x speedup over state-of-the-art compilers.

Successfully fuses diverse cascaded reduction workloads.

Matches the performance of hand-optimized kernels.

Abstract

Operator fusion, as a key performance optimization technique in the deployment of AI models, significantly improves execution efficiency and has been widely adopted in modern AI compilers. However, for cascaded reduction operations involving multiple loops with inter-loop data dependencies, such as the safe softmax followed by GEMM within attention mechanisms, existing compilers lack effective automated fusion and kernel generation capabilities. Although some works have addressed specific instances through hand-crafted fusion strategies, their solutions are limited in generality and difficult to extend to other similar structures. Given the prevalence of such computational patterns in deep learning models, there remains significant untapped potential in achieving general and automated fusion optimization. In this paper, we present a formal theoretical methodology for analyzing cascaded reductions which can fuse them into a single loop and introduce an incremental computation form. Based on this methodology, we design Reduction Fuser (RedFuser), a framework that automatically identifies supported cascaded reduction patterns and generates optimized fused kernels. Experiments show that RedFuser successfully fuses diverse workloads, achieving up to 2$\times$ to 5$\times$ speedup over state-of-the-art AI compilers and matching the performance of highly optimized hand-written kernels. The code is available at https://github.com/alibaba/redfuser