Asynchronous distributed-memory task-parallel algorithm for compressible flows on unstructured 3D Eulerian grids

•.A finite element method for the simulation of compressible flows using the Charm++ runtime system has been implemented.•Strong and weak scalability up to and computational cells, respectively, have been demonstrated.•The benefits of automatic load balancing in Charm++ have been demonstrated.•The f...

Full description

Saved in:
Bibliographic Details
Published in:Advances in engineering software (1992) Vol. 160; p. 102962
Main Authors: Bakosi, J., Bird, R., Gonzalez, F., Junghans, C., Li, W., Luo, H., Pandare, A., Waltz, J.
Format: Journal Article
Language:English
Published: United States Elsevier Ltd 01.10.2021
Elsevier
Subjects:
Automatic load balancing
Charm
ENGINEERING
Finite element method
Flux-corrected transport
MATHEMATICS AND COMPUTING
Shock hydrodynamics
Finite element method
Flux-corrected transport
Automatic load balancing
Shock hydrodynamics
Charm
ISSN:0965-9978
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•.A finite element method for the simulation of compressible flows using the Charm++ runtime system has been implemented.•Strong and weak scalability up to and computational cells, respectively, have been demonstrated.•The benefits of automatic load balancing in Charm++ have been demonstrated.•The full source code is available at quinoacomputing.org. We discuss the implementation of a finite element method, used to numerically solve the Euler equations of compressible flows, using an asynchronous runtime system (RTS). The algorithm is implemented for distributed-memory machines, using stationary unstructured 3D meshes, combining data-, and task-parallelism on top of the Charm++ RTS. Charm++’s execution model is asynchronous by default, allowing arbitrary overlap of computation and communication. Task-parallelism allows scheduling parts of an algorithm independently of, or dependent on, each other. Built-in automatic load balancing enables continuous redistribution of computational load by migration of work units based on real-time CPU load measurement. The RTS also features automatic checkpointing, fault tolerance, resilience against hardware failure, and supports power-, and energy-aware computation. We demonstrate scalability up to 25×109 cells at O(104) compute cores and the benefits of automatic load balancing for irregular workloads. The full source code with documentation is available at https://quinoacomputing.org.
Bibliography:USDOE Laboratory Directed Research and Development (LDRD) Program
89233218CNA000001; LA-UR-20-21450; LDRD-20170127-ER
LA-UR-20-21450; LDRD-20170127-ER
ISSN:0965-9978
DOI:10.1016/j.advengsoft.2020.102962