I've integrated fluids into my physics engine. Fluids are made of the same particles as solid matter, but computed differently. Phase change is seamless, matter melts and hardens back on temperature change. It runs fast, and world size limits are wider than any game would need.

You have rediscovered Clonk (and… well, I’m grateful for that)

I'm actually enchanted by Clonk design and mechanics and the game I'll be making might be heavily inspired by it. Though, it's not decided for sure.

Please do!

Finally someone will make that stupid mobile game ad about lava, gold and water a reality.

Noita 2 confirmed?

Is it your own engine or it runs inside a regular game engine? It's absolutely amazing.

It runs inside Unity. Thank you!

Is this something you're planning to release as a plugin? I'd pay quite a lot to have access to this kind of technology. Ideas for games that utilises this just kills my imagination everyday I see it 👀🥹🥰

I might release it, yes. After I make a game with it. There's so much to improve until it's really useful and convenient to make games with.

Yeah I totally get you. And How's the simulario going? Are you planning to release Beta / demo? It clicks right buttons in me head, can't wait to give it a go 🙈

Simulario has been in R&D phase for many months, and finally I've implemented all physics features I needed. I'm currently finalizing the physics, to move to rendering for a few improvements, and then will go straight to implementing actual game design.

Sweet. I hope it will go smooth from now on. Keep us posted 🔥

Veeery satsifying to watch already.

Looks awesome! Is it still Lennard-Jones or is it completely different kind of simulation? By the way do you use the standard 12-6 Lennard-Jones potential, or some simplified version of it, like just some linear attraction–repulsion (which technically isn’t Lennard-Jones)?

It's a blend between solid material (LJ), and PIC/FLIP fluid simulation. LJ is 13-7. I don't remember why I chose it and not the default 12-6. Using linear attraction-repulsion wouldn't let particles form such stable and solid material. Though, even LJ is not ideal for 2d. It was made for 3D simulations after all.

Wow! I was sure you were using the term “Lennard-Jones” loosely. I was under the impression it was too stiff with such large exponents for real-time simulations. I also thought you were using springs to make materials solid. How many sim steps/iterations per frame do you run?

16 a frame, dt is really small. But it works solid 150 fps on my 3090

16 is smaller than I expected, but I guess that’s because it’s running unconstrained at 150 fps? So basically we can say that "effective" iteration count per frame on 60 fps display/video is 40 (16 x 150 / 60)? i.e. dt should be 1/150/16 = 1/60/40 = 0.0004166666667

Almost correct, dt is 0.0002. When I make it 0.0004, simulation explodes

Is the physics really running on GPU (as your bio suggests?). Looks amazing.

Correct, runs on GPU. Thank you.

I imagine when you say "any game" you're not including something like Terraria, right? If that sort of scale isn't viable out of the box, I imagine it could be chunked or something.

Here the size is shown. Camera has moved to the right by maybe 10% of world size in this video. Same length goes underground. So, plenty of space.

@CritterObserves Is this real time? I hit massive roadblocks with liquid physics when it comes to this stuff which turned me off it, as the visuals aren't 3090 worthy but require a 3090.

@CritterObserves There's plenty of optimization possibilities, so it's totally doable running a good quality particle simulation with tens of thousands interacting particles on something like 3070, with cute enough rendering. If you go pure fluid, it's even faster, as it allows 10x time step.

For myself specifically I was looking at Noita style game with the addition of somewhat realistic physics and first handful of options I looked into didn't seem do-able at a convenient enough efficiency (for me on a 3060). Fractala seems to have it though, and some of yours as well. I'm still wondering if its functional enough to be the focus of a full game and not in like a sandbox style one.

@CritterObserves somewhat realistic liquid* physics.

@CritterObserves But it sounds to me like I missed a lot of options.

@CritterObserves I think it's doable,and you can do it if experiment a bit more with different approaches and optimisation tecniques.

Fluids are cool enough as is. Seamlessly transitioning between fluid and solid is insanely cool!

Needless to say, this is absolutely AWESOME. Keep up the great work!

so smmooooooth! remarkable!

Interesting! Reminds me of desiccant gels when they've absorbed water. Looks great

Cool How does this run fast? Any articles or Tutorials to help? I feel you aren't using water particles and apply physics on each particles like my dumb brain would

Fluid sim uses grid, where it solves incompressibility. Running each cell of the grid would be slow, because there's too many. So, I made it so only cells that have partciles would be computed. This way I have both huge world and fluid is computed fast.

Ohh I get it now ,cool so the entire mesh is made into grids... How do you handle physics tho any tips for fluid physics apart from studying 300paged of lecture material🤧🤲

PIC/FLIP method for fluids is quite simple, you just put particle velocities to grid, solve incompressibility, and return velocities to particles. It all boils down to adding a few numbers a few times, but would still require reading a few pages of explanations.

Ohh thnx , I do understand the principle from what explained tho Calculations are not done on particles directly but are applied instead... Cool stuff

For the fluid it's indeed calculations on the grid. But for the solid part it's each particle computes forces between its neighbors, so per particle calculations.

Ohhh ok ok makes sense, thanks alot❤️

This is beautiful! 🤩 Suggestion: rather than having the temperature, decrease linearly, use exponential decay. Then, dissipate heat based on surroundings, to allow hot particles to stay hot if they are near other hot particles, so that large clumps cool slower.

It's an interesting and physical concept. If it's not much a performance burden, I might implement it. It could even be used for some gameplay mechanics.

Cool. Dissipation is probably likely to be somewhat computationally expensive, although perhaps not marginally so... but the exponential decay ought to be cheap, if you choose to implement that, since it's just a remapping of the linear elapsed time. 🤓

That's so cool! How are collisions for this many particles computed efficiently? Is it a delaunay/voronoi graph?

@f_allat There's a grid, that knows local particles, so each particle only interacts with few neighbors, therefore, it's about O(k * N) complexity.

It makes sense. Thank you!

>wider than any game would need

Looks nice.

Holy shit that's insanely fucking cool. I've been trying to imagine this concept in my head for a while to eventually implement it for like a dynamicly generated game mechanics, and you've represented it so perfectly. This fills me with unimaginable joy

I'm glad it gave you these feelings!

This reminds me of Noita

Can it do live multiplayer?

Currently no, as physic computation is not deterministic. But I'll try to fix it.

I'm seeing Liero and Scorched Earth possibilities all over this :D

This looks great. Could be used to make some really good Noita-likes

A lemmings reboot is in order.

This is EXCELLENT.

Worms Armageddon (1999) (but cooler)

looks a little bit like noita

Very impressive! especially the huge world from the other clip. Is this a standalone engine, or to be plugged into an existing engine via its computer shader api etc.?

It all runs under Unity. When I call it "my physics engine", I mean I created all physics from scratch, and it allows me building a physicsbased game on top of it.

Ooh awesome o: Hope you plan to release it as an asset one day :) (don't have plans for it myself I think, unless it could do what FluXY does in limited range, but there'd surely be demand)

Curious about the performance. Does it run smoothly (or at all) on mobile?

Performance depends on amount of particles and other quantity limits. It can run on mobiles with fewer particles. Maybe up to 10K. At some point I might make a mobile game with these physics.

Thanks for the explanation, that's very interesting! I think one concern is the huge array and operating on it on every frame. I don't think 100x100 would be enough for games though. If it's possible to load and unload groups of 10k particles at a time, that completes the puzzle

Noita 2 when looks amazing

This looks INSANE

Looks like Powder Games

It was the main inspiration.

time for some orbital mechanics? (I'm saying I would love to watch planets collide in this engine)

oh yeah, I planned making a simulation like this with massive planets colliding. Now with fluid dynamics it shall be epic.

If this was Minecraft 2.0 I would play it

Extremely Noita-coded. Hoping to see a cool sandbox game with this tech!