Scientific Applications:

CSE - Computational Science and Engineering Applications

Several significant engineering simulation applications use Charm++. We survey some of these applications below.

Profesor P. Geubelle and S. Breitenfeld of the Computational Solid Mechanics Group have developed CrackProp, an explicit Finite Element method simulation of viscoelastic and plasto-elastic mechanics. The simulation tracks conventional volumentric elements, coupled by flat "cohesive" interface elements using the Charm++ Finite Element Framework.

Dendritic growth

Professor J. Dantzig and Jun-Ho Jeong of the Solidification Processing Lab have parallelized their dendritic growth metal solidification application using the Charm++ Finite Element Framework. This adaptive mesh, implicit solver fluid dynamics application is quite different from the explicit structures codes normally used.

The framework handles the changes to the adaptive mesh by re-assembling the parallel mesh, repartitioning, and redistributing the mesh pieces. Since the mesh changes only rarely, this does not significantly impact the speed of the simulation. The implicit solvers are implemented using the conjugate gradient method, which solves a global matrix using local operations.

Profesor P. Geubelle and S. Breitenfeld of the Computational Solid Mechanics Group have developed CrackProp, an explicit Finite Element method simulation of viscoelastic and plasto-elastic mechanics. The simulation tracks conventional volumentric elements, coupled by flat "cohesive" interface elements using the Charm++ Finite Element Framework.

Dendritic growth

Professor J. Dantzig and Jun-Ho Jeong of the Solidification Processing Lab have parallelized their dendritic growth metal solidification application using the Charm++ Finite Element Framework. This adaptive mesh, implicit solver fluid dynamics application is quite different from the explicit structures codes normally used.

The framework handles the changes to the adaptive mesh by re-assembling the parallel mesh, repartitioning, and redistributing the mesh pieces. Since the mesh changes only rarely, this does not significantly impact the speed of the simulation. The implicit solvers are implemented using the conjugate gradient method, which solves a global matrix using local operations.

People

Papers/Talks

14-34

2015

[Paper]

[Paper]

Split-and-Merge Method for Accelerating Convergence of Stochastic Linear Programs [ICORES 2015]

12-02

2012

[Paper]

[Paper]

A Parallel Algorithm for 3-D Particle Tracking and Lagrangian Trajectory Reconstruction [Journal of Measurement Science and Technology 2012]

05-08

2005

[Paper]

[Paper]

An Integration Framework for Simulations of Solid Rocket Motors [AIAAPropulsion 2005]

00-04

2000

[Paper]

[Paper]

A New Approach to Software Integration Frameworks for Multi-Physics Simulation Codes [IFIP Conference on the Architecture of Scientific Software 2000]

00-01

2000

[Paper]

[Paper]

A Parallel Framework for Explicit FEM [HiPC 2000]

98-09

1998

[Paper]

[Paper]

Programming Languages for CSE: The State of the Art [IEEE Computational Science and Engineering 1998]

92-04

1992

[Paper]

[Paper]

Unsteady Fluid Flow Calculations Using a Machine Independent ParallelProgramming Environment [ParCFD 1992]