Localizing and Explaining Performance Defects in Heterogeneous Applications

Konstantinos Parasyris | 23-ERD-022

Executive Summary

We will develop an automated approach for detecting and fixing the root cause of computational performance defects both during the development process, for new applications, and post development, for existing ones, using sophisticated profiling and code mutation techniques. Detecting and correcting performance defects will improve the performance and portability of broad ranges of high performance computing applications.

Publications, Presentations, and Patents

K. Parasyris, G. Georgakoudis, E. Rangel, I. Laguna, J. Doerfert, "Scalable Tuning of (OpenMP) GPU Applications via Kernel Record and Replay" (Presentation, International Conference for High Performance Computing, Networking, Storage and Analysis, Denver, CO, 2023).