Intel and AMD merge PyTorch CPU optimizations into oneDNN library

What is changing

Intel and AMD are integrating their CPU optimizations for PyTorch into a unified library, giving developers tuning for both AMD and Intel chips through oneDNN. AMD has maintained a downstream version of oneDNN called ZenDNN, and that work is now moving into the open-source oneDNN project published by the UXL Foundation and driven by Intel.

AMD also announced that it is joining the UXL Foundation as a steering member. The company said the role influences the future of the foundation in exchange for a contribution of $20,000 per year.

Why it matters for developers

The oneDNN backend is used for CPU inference and is built to accelerate that workload for the latest Intel and now AMD chips in frameworks like PyTorch. When oneDNN was added to PyTorch as a backend in late 2023, the companies said it showed a 1.24x improvement with the Float32 data type and 2.21x for the BFloat16 data type.

According to the oneDNN GitHub, the library detects hardware features at runtime to generate optimized kernels for the platform, provides low-precision data types, and includes optimized instructions for common AI and machine learning operations such as matrix multiplication and convolution. For developers, the move simplifies working with PyTorch across different systems by ensuring a single backend provides the proper optimizations for the hardware being used.

EAG backing and wider support

AMD says the unification under oneDNN was recommended by the x86 Ecosystem Advisory Group, a joint initiative that primarily exists between Intel and AMD and was established in 2024. The two companies have collaborated through the EAG before, including on the ACE extensions for x86 described earlier this year.

Although the announcement is focused on x86, oneDNN is a cross-platform library that also works on Arm and RISC-V architectures. Intel originally developed oneDNN from previously existing deep learning libraries before it was folded into the UXL Foundation. Intel was among the founding members of the UXL Foundation alongside Arm, Qualcomm, Google Cloud and others.

Sources: Tom's Hardware