To simulate an AI Engine kernel you need to write an AI Engine graph application that consists of a data-flow graph specification which is written in C++. This graph will contain just the AI Engine kernel, with test bench data being provided as input(s) to the kernel. The data output(s) from the kernel can be captured as the simulation output and compared against golden reference data. This specification can be compiled and executed using the AI Engine compiler. The application can be simulated using the AI Engine simulator. For additional details on the simulator, see Simulating an AI Engine Graph Application in the AI Engine Tools and Flows User Guide (UG1076).