My next video will actually be more difficult then I thought. I initially wrote "New Armstrong", but virtually nothing is known about that... I think I'll have to do Blue Origin in general...

@uhoh I started figuring out forces :D! Here's the BarnesHut n-body algorithm: imgur.com/a/Qi1rmaN

@MagicOctopusUrn I don't see any evidence of interaction. They all appear to just go around the central body. Gravitational Interaction would at least occasionally result in some large angle scattering - thinks kicked into eccentric orbits or even hyperbolic escape orbits.

So even if it's calculating interactions, they are not at least strikingly obvious.

Also, as soon as whatever that was finished I was directed to what looks like a creepy advertisement and I don't like that. Why did that happen???

In fact, if you did that on purpose I think it's quite naughty. I won't be clicking any more of your links. Ever.

Why does a marketing company's url generate what looks like a google page??

It has stopped now, strange (see comments)

@uhoh - that sort of thing is not uncommon - compromised advertisement slot is most likely

@RoryAlsop thanks for the feedback. It's never happened to me before. I didn't realize redirects were still so easy to pull off, and that imgur would be unable to screen or prevent it. Oh well.

So I think I will only click on i.stack.imgur.com/xxx.jpg type links from now on, and not raw imgur links. If it's not YouTube or one of those, then I won't look at it.

@uhoh tools like ghostery, ublock or scriptsafe are pretty good at preventing things like that, but for sites that do require a bunch of things to be loaded they can cause extra effort to just browse

@uhoh oh the mass of the object isn't relative to their size yet heh. They do interact with each other but it's just the BarnesHut algorithm applied on each iteratively. Assign each body a quadrant, apply all force in terms of that quadrant, repeat. imgur.com/a/Qi1rmaN shows instabilities, the one I linked had also already been running for awhile and stabilized.

Wait no, that one was wrong too, had a bug that was only performing it on the sun, good catch. i.imgur.com/BtfKp9O.gifv

Also, @uhoh it's probably something with Imgur... they have weird ads on videos, no ads if using ad blocker though. They only happen on video gifs of longer than 10 seconds

@MagicOctopusUrn okay well I've had enough "trauma" for one day. Personally I like keeping my simulations in script (e.g. python) so I can just run them without relying on the big bad internet.

Have fun!

@uhoh fair enough hah. I'll have to make my own GIF exporter, sounds easy enough.

@RoryAlsop yes I remember ghostery from years ago, thanks.

@MagicOctopusUrn I just looked up en.wikipedia.org/wiki/Barnes%E2%80%93Hut_simulation

@uhoh is it a good approximation?

That's really not the most appropriate way to approach a solar system simulator. It's important when simulations involve incredibly huge distances, medium distances and very very small distances all at the same time. That's why they use it for galaxies.

For a solar system simulator with say hundreds or thousands, but not millions or billions of objects, just use straight n-body, everybody talks to everybody, with no clever tricks.

These days laptops are gigaflops, even without GPU help.

@uhoh yeah, I watched a video on one run with like a million particles, they added collision and ensued conservation of mass. The way it made clusters about something with a mass of a super black hole was neato.

They used a much better scaling for masses than I am, I honestly just wanted to try something with forces over time lol

Even a Raspberry Pi can be 100 Megaflops

Jeez, that's awesome, I wonder what a cluster of 1,000 raspberry pis could do

What's wrong with that pastebin example I gave? That's got everything you need.

assuming you are doing like the solar system, planets, moons, etc.

@Uhoh it's great, but I lack the basic understanding of discrete mathematics to be honest, I want to understand how derivatives are really supposed to work over time first. I tried ODEs but it was a bit beyond me. This is more simple (imo).

If it actually runs out of steam at some point, then you can slowly ease your way into the fancy stuff

okay, so you are using an existing routine then? Because of course an orbit simulation with Barnes–Hut still has an ODE simulator, maybe you just can't see it. Are you running some package?

Nah, it's step-based integration. I iterate across all bodies, the algorithm I'm using is O(n^2) to avoid doing derivatives.

I'm starting to understand why an ODE is important to efficiency.

(And probably accuracy)

I don't know what that means exactly. you have a force, that gives you an acceleration, how do you exactly get the new position and new velocity?

can you show the few lines of code that do that?

Yeah hang on

github.com/carusocomputing/Barnes-Hut-Rewrite/tree/master

github.com/carusocomputing/Barnes-Hut-Rewrite/blob/master/src/…

public void start(), public void addforces(Quad q, int N)

Quad is a quadrant, you iterate through each body, applying forces in a quadrant around each body. I think the scaling is severely off, or something is severely off.

I honestly have expected to get a message from someone at some point "this isn't how physics works, none of this is how physics works".

Okay, this isn't how...

Line 140-ish

// Calculate the new positions on a time step dt (1e11 here)

bodies[i].update(10 * 1e11 / t.getValue());

well I can't help more. This is probably not what you want.

@uhoh Isn't how to what? I thought that would just increase the dt's scaling from the slider in accordance with steps to make the simulation go faster or slower.

I'm not really going for planets, I just want to play with mass at the moment and understand physics better!

okay, I can't help further then.

@MagicOctopusUrn no matter where you ask, step 1 will be to solve a differential equation. It's not like there is any alternative. I can not understand what you think you're going to find besides that to solve a multi-body problem. If you find something besides solving a differential equation, that will be profound.