Difference between revisions of "SC²S Colloquium - March 10 2017"

From Sccswiki
Jump to navigation Jump to search
 
Line 13: Line 13:
  
 
== Andreas Molzer: Optimization of a solver for the shallow water equations for heterogeneous GPU architectures==
 
== Andreas Molzer: Optimization of a solver for the shallow water equations for heterogeneous GPU architectures==
TBA
+
This work builds on past endeavours to implement an augmented Riemann solver for the Shallow Water Equations. The solver itself is based on work by D.L George and will be included without many changes. This paper deals with general concepts for implementing optimized algorithms for GPGPUs and show their application by constructing a better implementation of the previous \cuda solver code. The SWE-X10 project, aiming at providing a highly scalable simulation on heterogeneous systems is used as a framework. Additionally to dealing with \cuda, some general improvements are implemented in the surrounding project to bring it closer to its goal.
 +
Lastly, the solver for graphics cards is examined in the context of the project by finding generally good scalability properties.
  
  
 
[[Category:ShowComingUp]]
 
[[Category:ShowComingUp]]
 
[[Category:news]]
 
[[Category:news]]

Latest revision as of 13:34, 16 February 2017

Date: March 10, 2017
Room: 02.07.023
Time: 3:00 pm, s.t.

Andreas Schmelz: Optimizing Molecular Dynamics Simulations Using One-Sided MPI-Communication

TBA

Andreas Molzer: Optimization of a solver for the shallow water equations for heterogeneous GPU architectures

This work builds on past endeavours to implement an augmented Riemann solver for the Shallow Water Equations. The solver itself is based on work by D.L George and will be included without many changes. This paper deals with general concepts for implementing optimized algorithms for GPGPUs and show their application by constructing a better implementation of the previous \cuda solver code. The SWE-X10 project, aiming at providing a highly scalable simulation on heterogeneous systems is used as a framework. Additionally to dealing with \cuda, some general improvements are implemented in the surrounding project to bring it closer to its goal. Lastly, the solver for graphics cards is examined in the context of the project by finding generally good scalability properties.