Investigate why DSST propagator is slow with high degree and order of geopotential terms
I'm comparing DSST and numerical propagator performance in a GEO context:
- Solar radiation pressure
- luni-solar attractions
- Geo potential terms
Propagation time is 7 days. DSST initial state is osculating and propagation type is also osculating. I'm using Orekit 11.3.2.
The configuration file: inputs.yaml
Using 20x20 geopotential terms the comparison in terms of computation time gives (DSST and numerical have similar computation time):
- Numerical wall clock run time (s): 2.573s
- DSST wall clock run time (s): 2.568s
Using 6x6 geopotential terms the comparison is: (DSST is about 4 times faster)
- Numerical wall clock run time (s): 2.21s
- DSST wall clock run time (s): 0.6s
I'm not surprised by the second results (i.e., using 6x6). But I think we shall investigate why DSST computation time increase 4 times by switching from 6x6 to 20x20. At the opposite, numerical propagation only increase by 0.3 seconds.
In that case, we shall probably use Yourkit tool.
Edited by Bryan Cazabonne