HI!
We are preparing to develop an NPU. If possible, we hope to use OpenVX to complete optimization and neural network deployment. I’m sorry, I haven’t found any information on how to integrate a new hardware component into OpenVX, how to define the binary interface for hardware-software interaction, and whether there is any documentation or reference implementations available.
Additionally, as a beginner, I have another question: What are the advantages of OpenVX compared to MLIR and TVM?