Difference between revisions of "Essense Code Optimisation"
(Created page with "{{project info |description=Essense Code Analysis and Optimisation <!--|research project=--> |fields=Computational science; Computational fluid dynamics |financing=SNIC |active=y...") |
|||
| Line 9: | Line 9: | ||
}} | }} | ||
| − | Essense is a CFD code which | + | Essense is a CFD code which is characterised by summation-by-parts operators in the middle of the high order finite difference method, and weak boundary conditions. This project aims for "a better understanding of the bottlenecks of this code and learning about performance optimization potentials" (1st Phase for 1 person month) and "obtaining a revised code prepared for fine-level optimizations" (2nd Phase for 2 person months). Revision of the code includes the reduction of communication calls which is essential to achieving the better communication performance and the change of time-consuming loops for increasing the cache usage and improving vectorisation capability. In the long run, a speedup of this code will enable us to perform bigger and more detailed flow simulations within a varying range of scientific projects. The ambition is to prepare the code for PRACE Tier 0 type scaling. |
Latest revision as of 13:39, 15 October 2014
| Name | Essense Code Optimisation |
|---|---|
| Description | Essense Code Analysis and Optimisation |
| Project financing | SNIC |
| Is active | yes |
| Start date | 2014-08-01 |
| End date | 2014-12-31 |
Essense is a CFD code which is characterised by summation-by-parts operators in the middle of the high order finite difference method, and weak boundary conditions. This project aims for "a better understanding of the bottlenecks of this code and learning about performance optimization potentials" (1st Phase for 1 person month) and "obtaining a revised code prepared for fine-level optimizations" (2nd Phase for 2 person months). Revision of the code includes the reduction of communication calls which is essential to achieving the better communication performance and the change of time-consuming loops for increasing the cache usage and improving vectorisation capability. In the long run, a speedup of this code will enable us to perform bigger and more detailed flow simulations within a varying range of scientific projects. The ambition is to prepare the code for PRACE Tier 0 type scaling.
Members
| Centre | Role | Field | |
|---|---|---|---|
| Soon-Heum Ko (NSC) | NSC | Application expert | Computational fluid dynamics |
