This is my project repository for Hardware System Design course at Seoul National University, 2021 Spring semester.
The goal of my final project (which are inside the folders whose names start with "project_...") was to build a hardware accelerator that accelerates matrix-matrix multiplication operation. Along with building the hardware accelerator itself, I added additional features of 8-bit quantization, zero-skipping, and DMA module in order to speed up the operation even further. The resulting HW-SW system was tested on convolutional and dense layers, and attained up to 2x speedup compared to pure software implementation. The implementation was done with Verilog and C++, and deployed on ZedBoard.
Final Grade : A+