Abstract
This paper presents our work to bring state-of-the-art shared-memory programming to granular force kernels in the LIGGGHTS DEM code. A performance benchmark, based on miniMD, was created to simplify the task of testing and measuring performance impact of different parallelizations. Parallelizations which are found to have positive impact are planned to be brought back to the LIGGGHTS code base.
| Original language | English |
|---|---|
| Title of host publication | 6th International Conference on Discrete Element Methods DEM6 |
| Editors | Mustoe, Graham |
| Publisher | Colorado School of Mines |
| Pages | 181-185 |
| Number of pages | 5 |
| ISBN (Print) | 0-918062-20-9 |
| Publication status | Published - Aug 2013 |
Fields of science
- 103032 Fluid mechanics
- 203 Mechanical Engineering
- 203016 Measurement engineering
- 203024 Thermodynamics
- 211104 Metallurgy
- 204006 Mechanical process engineering
- 204007 Thermal process engineering
- 103043 Computational physics
JKU Focus areas
- Computation in Informatics and Mathematics
- Mechatronics and Information Processing
- Engineering and Natural Sciences (in general)