Back to Projects
Systolic Array-based Matrix Multiplier for Neural Networks

Systolic Array-based Matrix Multiplier for Neural Networks

Systolic Array-based Matrix Multiplier for Neural Networks (very in-progress)

Dec 2025 - Present 1 month Ongoing

Tech Stack

SystemVerilogVivadoComputer ArchitectureSystolic ArraysAXI-StreamQuestaSimPython

About This Project

This project implements a weight-stationary 8x8 systolic array matrix multiplier designed for neural network acceleration. It currently has a complete datapath with a controller for computation, with current plans to implement tiling to allow for any-sized matrix multiplication. It currently supports the INT8 data-type and is written in SystemVerilog.

Features

Weight-Stationary Systolic Array Architecture

Skew Buffers

Verification Methodology

Testbenches!

Golden Models

In-Progress Improvements and Future Plans

Tiling

View All Projects