AMD GPUOpen recently introduced MinDXNN, a new native HLSL and DirectX 12 library designed for lightning-fast Multi-Layer Perceptron (MLP) inference. The library targets AMD Radeon RX 9000 series graphics cards, leveraging their dedicated matrix cores through cooperative vector APIs to enhance performance. AMD states MinDXNN provides optimized kernels, sample code, and full source documentation "to remove compute interop friction," aiming to streamline the development of AI and machine learning workloads directly on DirectX 12 platforms. This initiative facilitates closer integration of high-performance AI inference capabilities within the DirectX 12 ecosystem on supported AMD hardware.