Galaxy collision in 3D, first attempt, 131,000 particles
Here's my first test of 3D astrophysics simulation.
There are 131,073 particles, with about 100 mass particles.That is to say about 130,900 massless particles. Gravity and time integration are computed on GPU using OpenCL. With such a small number of mass particles, barnes hut in slower than straight gravitation computation.
Much work has been done in user interface for 3D editing, which can't be seen on this video. Now, improvements are going to be done in camera rotations and code cleaning.
Still no SPH implemented, already tested it, it's very too expensive for real time when used with gravity computation...
For time integration I use 1rst order Euler integration. Before I used 2nd order leap frog integration, but I switched recently to 1st order Euler one, because strangely it appears to be more precise : when a galaxy is set up and animation starts, leap frog make center particles to go away, whereas Euler integration keep everything in place. In both case, particles are instanciated with an angular velocity corresponding to circular orbit.
Stay tuned
~~~~~~~~~~~~~
| VIDEO LOG |
~~~~~~~~~~~~~
Number of particles : 131073
Max tree depth : 256
Gravitational approximation : 0.10
Gravity : 0.000100
Gravity smoothing : 0.001000
Particle site : 62.765625.0
Particle factor : 19.259260.0
Lowest acceleration : 0.000000.3
Highest acceleration : 0.050000.3
Time step : 0.010s
Simulation time : 4m 21s
Simulated time : 1m 0s
Video time : 20s
Playback time scale : 3.00x
Simulation factor : 0.229885
Frame per seconds : 23.0fps