ML Based Co-processor Verification in SoC Environment

Abstract

 A co-processor based on machine learning is a highly efficient parallel compute block that is especially intended for the efficient execution of complicated computations such as neural networks. This project's CNN-based ML co-processor uses it to do calculations more quickly. This co-processor takes the input of Data, weights and the biases in order to provide the output results by performing the machine learning operations such as Data Canvas, Quantization, Convolution, Padding, Stride, Pooling and Activation functions. The main reason for dedicated processors for machine learning is enhanced energy efficiency which provides faster performance and reduction in model size and complexity. This paper involves the System on Chip(SoC) verification of ML based Co-processor using languages such as C and System Verilog in the C_UVM verification environment at the SoC level. The test cases are written using C and UVM for verifying the functionality of the design. The Functional Verification of the ML based co-processor tools is done using such as Vim Editor tool is used to write and edit all the test cases using the C and System Verilog. Arm Tool chain used to convert the C assembly code to binary format. The Cadence Incisive Tool used for simulation of the test cases and the Cadence Sim Vision to analyze the waveform that are simulated for the written test cases.