Seminar: Programming Heterogeneous Systems by Constructing Domain-Specific Source-to-Source Compilers
Peng Jiang
Assistant Professor, Computer Science
University of Iowa
Monday, February 24
9:30 - 10:30AM
1100 Torgersen Hall
Abstract
Modern computers have become heterogeneous by incorporating accelerators like GPUs. Writing efficient code for these systems is much harder than CPU programming. Traditional system software, including runtime libraries and general-purpose compilers, struggles to hide this programming complexity. A promising solution is to build domain-specific compilers that automatically generate and optimize code for accelerators. In this talk, I will present my recent work on domain-specific compilers for graph learning and mining applications. I will also demonstrate how our Cuke compiler, based on functional programming, provides a lightweight framework that simplifies domain-specific compiler construction.
Biography
Dr. Peng Jiang is an Assistant Professor in the Computer Science Department at the University of Iowa. Before joining UIowa, he obtained his PhD degree in Computer Science from the Ohio State University. His research focuses on High-Performance Computing and Compilers. His current work explores new programming techniques to simplify the development of high-performance and parallel programs for data science and machine learning applications. He is a recipient of the NSF CAREER Award for research in compiler and runtime support for heterogeneous systems.