Research Project Title:
Profiling Execution of Auto-Parallelized Programs in the Swarm Architecture
abstract:While the performance improvements of single-core processors have recently stagnated, the Swarm architecture has demonstrated novel advancements in the performance and scalability of program parallelization across a multi-core architecture. In my project, development of a generalized compiler for the architecture has begun, which will allow any program to automatically take advantage of these performance gains. The goal of my project is to design and implement tools to profile and parse data about how these programs are automatically parallelized and executed, such as how loop iterations are scheduled, or how function calling is organized, to better understand how a program is best converted into a Swarm program in order to maximize parallelism and performance.
“I am participating in this SuperUROP because of my interest in computer architecture and low-level software. I believe that this project is very exciting and at the forefront of development in these fields, and I feel my experience in these areas from classes such as 6.175 (Constructive Computer Architecture) will help me make a meaningful impact on the project. I also hope to gain valuable research experience and further my knowledge and understanding of computer architecture.”