Technology
AMD and Intel pool their x86 work for PyTorch in oneDNN
AMD will fold its ZenDNN optimisations into oneDNN, the open-source deep-learning library, under an agreement with Intel that both companies say gives developers a single place to find CPU acceleration for PyTorch across x86 chips. The step follows a recommendation from the x86 Ecosystem Advisory Group, and AMD has joined the UXL Foundation as a steering member. The stated aim is less software fragmentation.