Microchip's CoreVectorBlox IP is a neural network accelerator designed for their PolarFire FPGAs, enabling power-efficient AI/ML inference at the edge. It allows software developers to implement trained neural networks (from frameworks like TensorFlow or ONNX). The CoreVectorBlox IP itself comprises a Matrix processor (MXP) for general neural network layers and an MXP CNN IP specifically for convolutional layers. This architecture allows it to be instantiated as either a single core accelerator or in multi-core configurations to distribute neural network workloads. This makes it ideal for applications requiring low power consumption and small form factors.