As per my understanding, both TVM and MLIR are used as compiler infrastructure for deep learning neural networks. Is my understanding correct?.
And Which would be better if we are building a compiler for custom hardware that runs deep learning inferences?