AMD has officially joined the UXL Foundation as a steering member as of October 20, 2026. This transition elevates the company from its previous role as a general contributor to a leadership position within the organization.
A key outcome of this membership is a collaboration with Intel to establish oneDNN as the default open-source library for x86 CPU acceleration within the PyTorch framework. This integration aims to standardize the math operations performed during deep learning workloads.
The initiative seeks to minimize software fragmentation for x86 CPUs by reducing the need for developers to maintain separate code paths for different chip vendors. The UXL Foundation, established in September 2023 under the Linux Foundation, manages oneDNN as a core project.
While supporting this shared standard, AMD will continue to develop its proprietary ROCm software ecosystem and maintain its existing independent partnerships within the PyTorch community.
