[Open Source] TC Particles: millions of GPU particles

Boids? COOOOOOOL!!! Can't wait to get the new version in my hands.

 

Quite excited about that, as well as the new test scene !

 

I love you and I will buy you a hooker in Amsterdam.

 

+1 request for common particle types, especially snow.

 

For turbulence textures, how are the pixels packed? I have created my own vector field for a wind simulation, but I am unsure of how to pack this into a 3D texture for use as a turbulence texture. I've got it sort of working, but with a lot of fudging. Thanks!

 

Hi everyone, sorry for my absence Patch is becoming a little bigger - I need to refactor quite a bit of things, but it's coming along.

@sloopidoopi, ronan, Zen Davis: Thanks, I appreciate it! (Davis, I was in Amsterdam a few days ago, and I asked but nothing was arranged?)


@jerware: Good idea, we'll look into it

@valyard: 1. Certainly! That's exactly what a 'radial' force does. 2. Hmmm, theoretically, maybe. Currently it's not practical, and I don't see a widespread enough use for that for me to implement. Sorry!

@Pyromuffin: Make a turbulence force, and set it's mode to TurbulenceTexture. Then you can call from script: myTcForce.GenerateTexture(). It takes a 3D vector3 array. It has to be power of two size*. Let me know if that helps


*And in case you do want to pack it yourself: It's packed as an RGB value of normalized values. So: myVector * 2.0f - 0.5f; (which reminds me, that I really should handle vectors of a magnitude larger than 1 )

 

Inspiration : http://forum.unity3d.com/threads/179556-Particle-Rage

 

I am seeing some weird behavior when rotating turbulence textures. rotating 180 degrees by the respected axis causes the field to invert in that direction. This is pretty annoying because I'm trying to rotate the vector field in a room to cause the same effect, just mirrored. Various permutations of rotation and inverting the 'extent' properties can usually get this to work, but I wish the behavior wasn't so anomalous. Can you shed any light on how this works?

thanks!

edit: not to mention rotation of force objects is really weird, changing the x rotation by 180, also changes the z by 180 automagically(??), limiting which rotations are actually possible.

 

after much work, check out what I've done with TC particles and turbulence textures!

https://www.facebook.com/photo.php?v=4829671339220

 

The link isn't showing up, had to quote to find it out: https://www.facebook.com/photo.php?v=4829671339220

But that is awesome! Very curious about your setup for this, care to share? Kudos on the physical shaders as well btw

Also, I now see your previous post, not sure how I missed that. I'm not sure I understand the problem. Rotating the turbulence force does also rotate the direction of the force indeed. In technical terms, the vectors are in local space. Should that not be the case? Also, didn't know you could have negative extens to invert the texture, cool trick

 

Yes, I see now that it rotates each individual vector, rather than pivoting around a center point as it appears for the gizmo. Not sure whats wrong with the video, but I've taken out the video tags for now. Thanks, the physical based lighting with deferred rendering was really tricky, I know that n00body figured it out, I ended up using a duplicate camera that renders an extra gbuffer texture that contains physical properties, it's not perfect, but it works.

As for the wind simulation: At first I thought I could do this with the built in radial forces, but it turned out that it didn't look very good because the radial forces always also attracted toward the center of each door when it should be distributed more evenly around the area of the door. I actually fit a set of 16x16x16 cubes to each room and calculated the vectors pointing toward each door weighted by distance from each point on the door. This produces a vector field that gives each door's contribution to the air's motion. I modulate the force's power based on the difference in air densities between the two rooms when the door is open.

For some reason the extents have to be negative for this to work for me (and I'm using TCForce.generateTexture), I've checked the vector field and I'm sure it's correct. Even the gizmo appears to be backwards. I could send you a repro project if you'd like to see this behavior. I've also noticed that the turbulence textures seem to act in a spherical shape, despite being square shaped. Not sure what the intent is there, or if I'm just mistaken.

edit: I'm actually wondering how difficult you think this would be for me to implement a square vector field force if I were to purchase the source code version of TCparticles. $350 is a bit much for experimentation purposes- without knowledge of the architecture and especially without upgrade pricing, so I'd like your opinion.

 

Hi,
I played the demo and tried everything to force the fps goes lower but it didn't get lower than 49 with my GTX 580,so I liked this a lot.
But in the first scene the weapons no1 and 2 don't have a fade effect,was there any problem with this or we can figure it out using shader options?

 

Looking at the API, is there a way of getting the vector from each particle in some way? or set particle position? or is this handled by the gpu and there is manual access? cheers

 

What's the news on the next update? Waiting to hear from you as it's been a while since it was promised.

 

Hi everyone!

Back from a lovely vacation Back to work now!

@Pyro: Handled by email For the curious folks: Square turbulence forces aren't handled properly right now indeed. Turned out to be a bug!

@Hotsun: That's great! There was no fade, since the particles are opaque alpha cutout. Since they are not translucent, you can't really fade them. Additive particles can easily be faded though

@jomasan: Currently, no, but see patch notes below

@Tom Castle: I know, it has been way too long! My apologies for the delay, besie the vacation I've been working as hard as I can, but unfortunately mostly on some other stuff. Good news is, finally finished the patch today!

Patch 1.0.9
So, it had been hanging in the air for a loooong time, and it's finally here! It includes:

-Boids flocking, for lively firefly effects, or just cooler magic effects, whatever you want to do with it
-Overhaul of a big part of the structure - Greatly reduces CPU and RAM overhead for particle systems, especially in more complex systems. Additionaly, will allow for much easier patching and control in the future

-Fixed square turbulence forces
-Fixed turbulence forces not rotating properly
-Fixed mesh particle rotation in local space
-Fixed wireframe preview scale
-Fixed initial mesh burst on mesh emitter being wrong
-Bunch of performance optimizations

-Extension system


Extension systems? Bare with me. The idea is that you can write your own code to control the particles, quite low level, in a compute shader, after TC Particles has done all it's work. This means that you have fine control over each and every particle behaviour - Boids flocking could easily have been implemented as an extension method (might do so as an example at some point). I suck at explaining things, so I'd rather just show some code. I hope that makes it clear what I'm getting at:

Extension.cs:

Code (csharp):

1.  
2. public class ExtensionTemplate : MonoBehaviour
3. {
4.     public TCParticleSystem System;
5.     public ComputeShader Extension;
6.  
7.     [Range(0.0f, 10.0f)] public float AccelSpeed = 1.0f;
8.  
9.     // Update is called once per frame
10.     void Update ()
11.     {
12.                 //Bind own custom variables to compute shader
13.         Extension.SetFloat("AccelSpeed", AccelSpeed);
14.         //First bind the buffers
15.         System.Manager.BindParticleBufferToExtension(Extension, "MyExtensionKernel");
16.         //Then dispatch
17.         System.Manager.DispatchExtensionKernel(Extension, "MyExtensionKernel");
18.     }
19. }
20.  
21.  

Extension.compute:

Code (csharp):

1.  
2. #pragma kernel MyExtensionKernel
3.  
4. //*Obligatory defines go here, cut away for brevity*
5. float AccelSpeed; //set by CPU
6.  
7. [numthreads(TCGroupSize, 1, 1)]
8. void MyExtensionKernel(uint3 Gid : SV_GroupID, uint3 DTid : SV_DispatchThreadID, uint3 GTid : SV_GroupThreadID )
9. {
10.     //This example simply accelerates the particles
11.     SystemParameters syst = systemParameters[0]; //Always read the system first, handles bunch of stuff
12.     particles[GetId(DTid.x, syst)].velocity += AccelSpeed * syst.deltTime; //Note: Never read directly from DTid.x, but use GetId().
13.  
14.    if (particles[GetId(DTid.x, syst)].pos.x > 100.0f)//all particles that are too far out to the right
15.    {
16.        particles[GetId(DTid.x, syst)].life = 1.0f;  //kill kill kill kill!
17.    }
18. }
19.  

Add the component to some game object, link it all up, et voila, the particles of that system now follow these custom rules.

If you can write some compute shader code, this should make some sense to you. It's a really really basic system, but it does give low level access so for the hardcore coders. O, and did I mention all of this is in the indie version as well?

Patch has been submitted, should be live soon

 

Hey guys! I've been working with TC Particles a little and, as requested in this forum a few weeks back, I'm happy to share what I've managed to make with it. My studio's Kickstarter video showcases a few of the applications I've found for TC Particles so far, but as I get more familiar with it, I'm sure I'll find more. Thanks TCS for the great software! http://www.kickstarter.com/projects/1972129536/space-opera

 

That's great! I'm glad that you like TC Particles, good luck with the kickstarter Is it okay if I add this to the website for the list of projects using TC Particles? Thanks!

 

Can Tc particles do something like this? http://youtube.com/#/watch?v=RuZQpWo9Qhs

 

Hmm, not quite. You can make something that looks 'close enough', but it doesn't do actual fluid simulation. With some work you could make something that looks convincing though (and probably performs better anyway)

 

I've ran into a HUGE bummer of a bug.

Using the SSAOEffect on the camera with anything in the scene that has a transparent material, prevents TC Particles from rendering at all.

Any chance there's an easy fix for this?

 

It sounds like there used to be issues with the default Unity alpha-blended particles as well until an update to the SSAOEffect in 3.5. This somehow related?

 

Hi!

Well thats odd. I've been able to reproduce the bug so I should be able to fix it. I don't know much about the internals of SSAO, but I think I know what's going on.

As a temporary fix, open up the SSAOEffect.cs, go to line 87, and comment away the line [ImageEffectOpaque]. I will get back to you as soon as I have an actual fix

Edit: the plot thickens. I added the [ImageEffectOpaque] back in, and now it does work.... Seems like a bug in unity? Can you try this as well?

 

Thanks for looking into this!

Commenting out the line does fix it. When I added it back in though, it broke for me again. Not sure how it fixed itself for you. You must just be lucky

 

Do the particles always look like a bunch of little dots? Could this be used for 2.5D particle based water that respects gravity?

 

hi,

I just bought your asset from the asset store. Unfortunately I cannot get it to compile because of multiple errors in LibNoise. Perlin does not contain a definition for mutliple values in TCNoiseForceGenerator. Do you know how to solve this issue?

Edit: I think I solved it. Another Perlin.cs file was messing with TC Particles.

 

Wow, this is fantastic. I hope Unity hires you or buys this package and implements it into unity because this would really help it push to the next generation.

 

Well look at that! I go on a vacaion and unity puts TC particles on sale. Was a huge success! THank you everybody who joined the club

@krileon: Right now TC Particles works best for 'glowy stuff'. I am working on improving other render methods though, there;s already some beta stuff included in the package.

@CaptainChristian: Aaah, seems like I maybe should try to namespace that lib. Thanks for the heads up!

@Dooskington:

 

I am of those new club members, but unfortunately I have been quite unsuccessful in getting TC Particles to work. Currently I am just trying to run one of the test scenes, but all that happens in Unity is a black screen then a white Unity window with the error "Display driver stopped working and has recovered."

I am using Win 7 Professional 64-bit with a XFX Double-D FX-787A-CDBC Radeon HD 7870 GHz Edition Black Edition 2GB 256-bit GDDR5 video card and the latest (non-beta) driver from AMD.
I loaded into a brand new project under the latest free version of Unity.

Thanks for any help in getting this to work!

 

It's not stopped working but stopped responding, it's actually pretty important as that happens once the driver doesn't respond for a certain amount of time in recent windows OS, could be worth a shot for you to increase the (default 2 second) delay to a long time to see if it eventually works (meaning something is stalling the GPU for too long at once) or if it still doesn't (likely meaning the driver stoped responding forever and is hung)

You can try changing the registry this way :

1) Go to HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers
2) Right click on it choose new QWORD, name the key TdrDelay
3) Put the value 20 (it's in seconds, default is 2)
4) REBOOT => nothing happens until you reboot

After that re try what you did, if it works after a while it means it's just doing something too long at once, if it doesn't work / stay hung after 10+ seconds then there's definitely a problem.

Hope this helps

 

I did all of that, but I still am getting the same error (now it just takes a little longer). I also tried deleting the entire Unity project folder, rebooting, then creating just a TC Particles project again, but it didn't help.

BTW, thanks for the quick response!

 

Hey bryan! Crap, Im sorry to hear that. Thanks ronan for helping him out. Might be a good way to see how long your gfx card hangs. I think it might be an infinite loop though, so maybe that doesn't help.

Can you try making a simple setup? Just the default particle system. Then try some different emission shapes, and finally try emitting from a mesh and emitting mesh particles. Let me know what does and does not work. Thanks!

 

It's not an effect it's magic ò_Ó

 

It sounds like an infinite loop or race condition yes otherwise it would crash or at least respond after a while, 10seconds is a huge amount of time already (hell the default of 2 is pretty decent for a start).
A good try methinks would be (substantially) downgrading your driver, to something like a year back, see if it fixes it, if so it's a driver issue, if not then, good luck p v loon lol!

 

I have gone through all the shapes, mesh and emitting mesh particles - all worked within Unity. But while playing with the Inspector for TC Particles, I noticed something possibly weird:



Notice in the Particles section that Lifetime, Start Speed, etc do not have values and I cannot find a way to enter values for them. I went through your tutorial video and it looked like he didn't do anything special to modify these values...

Thanks,

Bryan

 

Hello,

is there a way to change the simulation space through script?

for instance:

particleSystemNice.GetComponent(TCParticleSystem).Emit (50);

works fine, but:

particleSystemNice.GetComponent(TCParticleSystem).SimulationSpace = "Parent";

is read only.


Thanks for any help!

 

@Korok: The biggest thing to happen to special FX since the invention of fire!

@ronan: Yeaah, this could be baad.

@Bryan: Aah, I see now. The problem is with unity 4.2 :S The annoying thing is, we noticed this in some other project in the 4.3 alpha and already sent them a mail about that. Turns out the same problem is in unity 4.2, I think we even notified them of that before the release of 4.2. Right now there really isn't a lot I can do about it, property drawers are just all kinds of broken. Luckily unity 4.2.1 is on the way, to fix this issue. But, If it takes them to long, I'll throw out the property drawers and do a fix that way.

@cmart: Aah, I see. Little messup in the API there. I'll expose it, no reason not to. Though, may I ask, what are you trying to do?

 

Thanks for looking into it! Basically, I have a "flaming sword" for lack of a better description, so I want to use the line emitter in parent space for say .1 seconds so it matches the angle/location of the sword, and then switch the space to world so the particles float off without being tethered to the movement of the parent. When the line emitter is set to anything but parent, it just emits in a straight line parrel to the floor, with no rotation. In parent mode, it will match its parents rotation, but then it will continue following around the parents transform and looks quite goofy : )




Thanks again for your awesome work! I absolutely love the particle system, I can't wait for more tutorial videos : )

 

Will 4.2.1 also fix my problem where I cannot run the demo scenes? Are these two problems related?

 

@p v loon : Hey, I just got my hands on the indie version. So far so good , except for the "property drawers" issue, i am in the same boat as Bryan Pope. But i have got one question, something that i don't see to do properly :
I have no issue to start the fx : " fxChaingunShoot.Play(true); " ... it starts properly.
But when i do the " fxChaingunShoot.Stop(true); " , it stops, but will never starts again. I put the " fxChaingunShoot.Play(true) " , but still stays quiet...

I thought it would be like the shuriken system, with simple play or stop commands, but seems like i am missing something here. Went through the documentation, tried to set DoEmit ( false ) instead, but it won t let me do it, says i can t use boolean on that. Any help would be appreciated.

EDIT : Okay, i was being dumb... only spent 4 hours chasing my own tail... i had a look at the sample scene with the gun and did " fxChaingunShoot.Emitter.DoEmit = false " or true...
Nevermind my question, leaving my stupidity here for others in case they need it.

 

@cmart: Aah I see! Hmm, hope that works actually. If you change it in the inspector, do the particles 'jump' ? If so, this needs a little more extensive solution to componsate for the change in space.

@Bryan pope: I don't think so :S Could you try doing the tests I described in unity 4.1?

@Feydwin: Sorry for not getting back to you earlier! But, glad to see you figured it out For the record, if you do want to freeze the particles, there is also a Pause() function.

 

Hello! I'm having the same problem with the inspector as Bryan. Is there any way we could get a hotfix? Maybe a workaround DLL without an inspector? I *need* something before the 7DFPS starts, because my team is going to make a game centered around particle effects and other pretty things! Please help, we really don't want to use Shruikan!

 

It seems to work well in the inspector, it is not "perfect" but I think it looks pretty good

 

Would also use the hotfix if possible. I am participating in VrJam and want to use TCParticles for my project.

 

Hello , this asset look awesome, I guess there no way that would work on mobile? Even with gles 3.0 ?

 

Hello everyone! Let me do some answers first, then a quick update on the business of things.

@CPTechnologies: Hotfix below
@cmart: Glad to hear! I'll send you a PM
@KarelA: Hotfix below
@toto2003: Thanks But unfortunately: indeed, not without major major refactoring

So, now for a quick update. I'm bad at these kind of things, let's do it Q&A style:

My inspector is broken!! Why?
Unity 4.2 breaks property drawers completely, at least for TC Particles. They just don't do anything. That's why you only get an ugly foldout.

Well, why aren't you fixing it? For that matter, where are the updates!?
22nd of july I went on my first big trip alone around the world, starting in vietnam. Currently I'm in Korea. Anywho, guess when 4.2 was released? 22nd of july, exactly. I only have a crappy laptop with me, and most of my wifi comes from coffee shops. Not to mention that I don't want to miss out on my vacation too much. It makes communication a bit harder, and doing proper fixes nigh impossible. I will be back early September. If you're really angry at me, you can give me a punch at unite I will be doing a talk on DX11 stuff there.

Alright, alright. So what now?
I'll try my best to support everyone individually, but pushing out an update will be hard! As to the problem with unity 4.2: I'm fortunate enough to have a job where I at least have some clue of what's going on in unity. We noticed property drawers being broken in the 4.3 alpha, and reported it to unity. I don't know what they did with it there, but they apparently didn't notice that the same thing goes for 4.2. We shouted at them a bit and they promised a quick fix in 4.2.1. For those of you who can life with it, there's a hackish hotfix down below so you can at least still sort of work.



Hotfix
Method 1: Go to Assets/Plugins/Editor. Locate TCEditorBase.cs and remove it. Replace with this file: http://g2f.nl/0d1a7yu
Method 2: Go to Assets/Plugins/Editor. Locate TCEditorBase.cs and open it. Change line 79 from:
EditorGUILayout.PropertyField(prop);
To:
EditorGUILayout.PropertyField(prop, true);

This will not bring back the property drawers, but at least make the foldout work. In the foldout, there are 3 values and a mode. Min and Max are used for the Between mode. Value will be used for the constant mode. This is not ideal of course! But maybe it is just enough to push you through the VRJam or 7DFPS until unity fixes the property drawers!

TC Particles really has had a major impact on my life, which feels really special. First it got me in touch with amazing people who I'm still friends with today, there was the DX11 comp, it got me a job offer working on an amazing game (more about that someday), and now the talk at unite! The financial side of things is. What's greatest of all though is seeing TCP be used for amazing stuff, and I hope I'll see more amazing stuff from you guys soon! Thank you all for the patience and amazing support

 

Thanks! Unfortunately the fix didn't work. However, I played around with TCEdtiorBase.cs, and found one more line that needs fixing to get all the properties in the particles tab visible and functional.
First, do p v loon's method 2, then go to line 85 and change it from
EditorGUILayout.PropertyField(CheckProp(varName), new GUIContent(dispName));
to
EditorGUILayout.PropertyField(CheckProp(varName), new GUIContent(dispName), true);

And it's fixed! It looks like it's completely functional, but I had *no* idea what it was like before 4.2, so don't go by me.

EDIT: Looks like forces are broken too! I'm looking for a fix.

MOAR EDIT: Found the fix! Go to Plugins/Editor/PropertyDrawers, find MinMaxDrawer, comment out line 5.
This may break some other things, but as far as I can tell, that class does nothing other than bug out due to Unity 4.2's property silliness.

 

Thanks pv loon and CPT.

loons solution worken only half way for me. Some properties were still locked. Added the stuff that CPT suggested and now everything seems to work OK.

Loon. Have a nice vavation and looking forward to hear yout talk at Unite.

 

Hey there,
Is there something special i need to do to make TCs work in webplayer? I've enabled DX11

 

@CPT: That is awesome! Thanks for finding that out. I'm really bummed that unite is taking so long with this 'hotfix' (more like a not fix, am I rite? ... I'll go now). In the meantime, here are the modified scripts to install the hotifx:

http://g2f.nl/0b5d4om

Backup your Plugins/Editor folder and replace with the one in the zip. Thank you all for the patience!



@KarelA: Great to hear that, community customer support, isn't that awesome? Looking forward to meeting you at Unite

@Pangamini: Shouldn't be. My guess: are you on nVidia optimus? Unite games don't trigger optimus, so you end up still using the integrated intel card, which doesn't do DX11, so it doesn't do TC Particles