Christophe Le Bris <chris.lebris@gmail.com> a écrit :
Hello,
Hi Christophe,
I have a question about a behavior that I observe when using the DSST propagator (from Matlab for my case). When I propagate an orbit with a NxN geopotential model, everything run ok and I get the results in a few seconds. But if I simply reset to the initialState and restart the same propagation, the run time increases (almost twice). And each time I do this? the run time increases. For example, a simple loop like this one: loop until xxx prop.setInitialState(initialState, false); finalState = prop.propagate(finalDate); end loop and each time, the execution time of the propagation increases. 5s - 9s - 13s - etc Am I the only one that observes this? Is there something that I do wrong? Any ideas? I have this problem only with DSST and not with the others propagators.
It is the first time someone notifies this behavior. Do you observe this with the recent 6.1 release or the older 6.0 one? We have been doing extensive simulations with DSST since a few months but these simulations were always in separate processes so its no surprise we have not seen it. We are also working on some Hansen coefficients caching which may improve the situation in the future.
From the behavior you describe, I would guess that something is stored in memory and looking back for it takes more and more time. Do you have a profiler installed on your system (I would suggest yourkit, which is a very good commercial tool) and could you look were processing time is concentrated?
Could you also set up a simple autonomous test case (as a Junit test if possible) showing the behavior so we can investigate on our side. Then you could open an issue on the forge and attach your test to it.
best regards, Luc
Thank you
---------------------------------------------------------------- This message was sent using IMP, the Internet Messaging Program.