1 Million particles N-Body Simulation

I finally got my Barnes-Hut program to a stage i can do big simulations in a reasonable time, so here it is. before this my biggest simulations were of around a quarter of a million particles, and with the brute force method i couldn't really go above that. This simulation does a million particles faster than my old program could do a quarter.

2^20 = 1048576 particles.
Calculated using the Barnes-Hut algorithm, which was programmed using c++ and Cuda.
Color in the simulation is based on speed.

The starting conditions are a disk spinning around a heavy particle in the middle.
The calculations and drawing took between 5.5 and 6.5 seconds per frame, coming to about ~10 hours for the whole video.
#simulation #programming #gravity

[jamie]

I really enjoy watching particle simulation. I'm going to sub and I hope you keep uploading simulations

[mahdiyarrazavi-ijhh]

Planet formation❌️
Galaxy formation✅️

[mrgoodpeople]

Need add dispertion for particle's velocity to prevent formation of globuls.

[ReaXTutorials]

And I don't even get any credit from this...

[oliviergiroud]

Have you tried to incorporate some dark matter hypothesis ? Or MOND gravity ? It would be interesting to see what fits the best with observations and what contributes to the quickest galaxy formatiib, to reflect latest James Webb observations

[zdlax]

I wonder if gravity sims would be much faster if the underlying models were general relativity and not Newtonian. Classical integrators scale as N^2 right? Wouldn't a GR based sim only need to calculate the contribution of each particle to the spacetime geometry, which I think is O(N) and then step each particle forward along the updated geometry, again even more certain is O(N)? Just a fleeting thought. Could be totally off.

[thesquee]

What exactly are you doing on the GPU? Any stuff with building / modifying the tree itself?

[anujchavda]

What is specifications of your computer/laptop

[desepticon]

Do the particles have collision, and thus a way to lose momentum?

[franckmalers]

Where is the code ! and also, with what do you generate the images? can it take hdf5 files as inputs?

[Francyrad]

Hello, why don't you implement your code in OpenSPH? In this way the code will work on multicore with collisions also

[Baconman]

I hope when this goes viral I will be at top comments 😃

[starpawsy]

Is that a 2D simulation or a 2D projection of 3D?

[anujchavda]

You are not doing this in normal specifications of the computer

[mrgoodpeople]

I make simulation with only 20K particles, but with friction for gaseous component. It's show how spiral galaxy may turn to barred galaxy:

And in this video I show how the gas convert to the stars and vice versa:

[jonathanbaincosmologyvideo]

So it does not satisfy the 'Rotation Curves of Galaxies Problem"?
Try with a binary system that emits stars...