Intel has long worked on tuning their oneDNN neural network library for their hardware and saw its adoption by PyTorch and other AI frameworks. AMD has maintained ZenDNN as their downstream fork that is oneDNN API compatible while catering to their Zen processor designs. As an interesting twist, AMD and Intel jointly announced today they will be collaborating around oneDNN for PyTorch and pushing the AMD Zen improvements into oneDNN.
0
The competing companies announced they will be working to combine their x86 CPU optimization work for PyTorch into oneDNN. This follows their work together on the ISA side with the x86 Ecosystem Advisory Group.
The two companies will focus on oneDNN as the default open-source library for x86 acceleration in PyTorch. AMD ZenDNN optimizations will go into oneDNN. Additionally, AMD will join the UXL Foundation as a steering member.
The UXL Foundation was formed in 2023 with many of the former oneAPI software projects going under the consortium backed by Intel, Samsung, Qualcomm, Arm, and other vendors. AMD is now joining that party in contributing to oneDNN.
Very interesting development and it wasn't on my bingo card this year seeing AMD contribute to oneDNN. More details via the AMD Newsroom.