Score: 2

Equivalence Checking of ML GPU Kernels

Published: November 16, 2025 | arXiv ID: 2511.12638v1

By: Kshitij Dubey , Benjamin Driscoll , Anjiang Wei and more

BigTech Affiliations: Stanford University Microsoft

Potential Business Impact:

Checks if computer code for AI works correctly.

Business Areas:
GPU Hardware

With the rapid progress of deep learning and large language models (LLMs), companies now spend enormous sums executing GPU kernels. These kernels have, therefore, become prime targets for aggressive optimization. Recent efforts increasingly leverage LLMs to generate GPU kernels, but make no formal guarantees about the generated kernels. We present the first equivalence checker for GPU kernels and use it to formally verify the correctness of machine learning (ML) kernels optimized by hand, by LLMs, and by compilers. We show that our equivalence checker is sound and, for a well-defined class of GPU kernels which includes the programs of interest, complete. Our implementation, VOLTA, can verify ML computations such as convolutions, matrix multiplications, and various attention mechanisms.

Country of Origin
🇺🇸 United States

Page Count
23 pages

Category
Computer Science:
Programming Languages