Upcoming Event: CSEM Student Forum

Asynchronous Tasks in Scientific Computing

Alexander Strack, University of Stuttgart

Abstract

Asynchronous Many-Task Runtimes (AMTRs) have emerged as a powerful alternative to established parallelization methods, such as fork-join and message passing. While the theoretical advantages of AMTRs are undeniable, determining how they translate into real-world scientific computing applications can be challenging.
In this talk, we introduce the concept of code parallelization using asynchronous tasks. We discuss the major advantages and drawbacks of this approach compared to traditional parallelization methods. Additionally, we present some of our results using the AMTR HPX for various applications commonly found in scientific computing. We also provide a brief overview of how developers can integrate different stages of HPX into their existing sequential or parallel code.

Biography

Alexander Strack is a third-year Ph.D. candidate at the University of Stuttgart. With a background in Computational Engineering, he has shifted his focus to High-Performance Computing. His research primarily concentrates on parallelization and optimization using AMTRs for scientific computing applications.