AMD and Intel have announced a collaboration to merge their x86 CPU optimization work for PyTorch into oneDNN.
oneDNN is an open-source deep learning library governed by the Unified Acceleration Foundation (UXL). The move follows a recommendation from the x86 Ecosystem Advisory Group.
One Library for x86 in PyTorch
Both companies have settled on oneDNN as the default open-source library for x86 acceleration within PyTorch. AMD is integrating its ZenDNN optimizations into oneDNN and has joined the UXL Foundation as a steering member.
The goal is a single place for developers to get CPU optimizations on AMD and Intel processors. That should make AI applications easier to build, maintain, and deploy across x86 systems, and reduce fragmentation in the ecosystem.
AI Compute Extensions (ACE)
The two companies are also working together on AI Compute Extensions (ACE), a next-generation instruction set architecture (ISA) for x86. ACE targets matrix multiplication, a core operation in AI workloads. It will reach PyTorch through future oneDNN updates.
Let me know if you want the headline or excerpt shorter, or the quotes trimmed.



