Scalable hierarchical high-order CFD solvers for future exascale architectures
Abstract/Contents
- Abstract
- With the public launch of supercomputer Frontier at Oak Ridge National Laboratory in May 2022, we are officially entering the exascale era of computing, characterized by systems capable of executing one exaFLOPS (10^18 floating-point operations per second). With the anticipated advent of additional exascale supercomputers in the forthcoming years, the extraordinary leap in computational power promises to revolutionize various fields for scientific discovery. For the Computational Fluid Dynamics (CFD) community to harness the power of these future supercomputers for complex Navier-Stokes computations, we must rethink and redesign numerical algorithms for extreme parallelism, paying particular attention to data placement and movement. In addition, advances in turbulence simulation technology will play a critical role in the analysis and design of aerospace systems. The accurate prediction of turbulence is essential for determining aerodynamic characteristics of aircrafts. High-order numerical methods have recently emerged as a promising approach to address various challenges of scale-resolving simulations. Yet, efficient preconditioners and linear solvers, which are often the most computationally intensive parts of implicit high-order codes, are still largely elusive. In this work, we aim to design a high-order CFD solver capable of providing high-fidelity solutions for scale-resolving simulations while simultaneously showcasing efficient scalability on prospective exascale supercomputers. To this end, this dissertation presents the development and implementation of a hierarchical scalable high-order CFD solver framework. The proposed framework integrates a promising hierarchical preconditioner based on Sparsified Nested Dissection and a communication-avoiding iterative Krylov solver with guaranteed numerical stability. The performance of the framework is evaluated in an implicit Discontinuous Galerkin high-order solver. With scaling studies, we demonstrate the algorithmic scalability and robustness of the overall framework for solving real-world aerospace engineering problems
Description
Type of resource | text |
---|---|
Form | electronic resource; remote; computer; online resource |
Extent | 1 online resource |
Place | California |
Place | [Stanford, California] |
Publisher | [Stanford University] |
Copyright date | 2023; ©2023 |
Publication date | 2023; 2023 |
Issuance | monographic |
Language | English |
Creators/Contributors
Author | Xu, Zan, (Researcher in exascale computing architectures) |
---|---|
Degree supervisor | Alonso, Juan José, 1968- |
Thesis advisor | Alonso, Juan José, 1968- |
Thesis advisor | Darve, Eric |
Thesis advisor | Saunders, Michael A |
Degree committee member | Darve, Eric |
Degree committee member | Saunders, Michael A |
Associated with | Stanford University, School of Engineering |
Associated with | Stanford University, Department of Aeronautics and Astronautics |
Subjects
Genre | Theses |
---|---|
Genre | Text |
Bibliographic information
Statement of responsibility | Zan Xu |
---|---|
Note | Submitted to the Department of Aeronautics and Astronautics |
Thesis | Thesis Ph.D. Stanford University 2023 |
Location | https://purl.stanford.edu/dq243nd9342 |
Access conditions
- Copyright
- © 2023 by Zan Xu
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...