Adjusting simulation timestep in Kangaroo2


(S T Molenaar) #1


Is it possible to adjust the timestep of the simulation in Kangaroo2?
Even with only a couple dozen points and springs it takes a considerable amount of time for the simulation to get to an equilibrium.
This is on a 6 core machine so it is not caused by a lack of compute power.


(Daniel Piker) #2

Can you post the file? With that many springs it should only be taking a few milliseconds

(S T Molenaar) #3

Example time (56.7 KB)

Try moving the slider at the bottom of the yellow section to compress the hypar structure.
The points get moved out of the block immediately but it takes some time for the structure to get to a stable state.

Thanks for helping!

(Daniel Piker) #4

I haven’t studied your definition in detail, but a couple of things I notice -

You have the energy threshold set to 1e-40
I don’t think there is ever any need to use anywhere near this low a value - the radius of an electron divided by the radius of the earth is several orders of magnitude bigger than that. The errors from the floating point operations will be bigger than that, so you are essentially guaranteeing the solver will never consider the result converged.

The other thing is the ratio of the strengths of the goals. The ratio of the strongest to weakest goals is very high - and this will slow down convergence.