Efficient data analysis is increasingly difficult as the growth of dataset sizes outpaces the growth of CPU processing power. For that reason, GPU-accelerated databases are an increasingly popular idea. However, two major issues that GPU-based systems face are data movement overhead and relatively expensive, limited memory. Our project focuses on solving both issues by implementing the delta binary packed encoding scheme for integer columns’ compression on GPUs, while assuming that the dataset is resident on the GPUs.

I am excited about SuperUROP because it lets me apply what I have learned from classes I have taken in previous terms, such as 6.824 (Distributed Computer Systems Engineering), 6.172 (Performance Engineering of Software Systems ), and 6.828 (Operating System Engineering), to a project with real-world load and resource constraints. This project will help me further understand the modern CPU/GPU architectures and how to program them efficiently.