New Fisix Demo - Blobs

I dont like to flood this forum with publicity for the fisix engine, but I feel that this demo is pretty cool :)

I've played gish many times and have tried to replicate the blobby type of creature many times in the past, and failed. After reading some threads on some forums today I found the solution... or at least it's getting close. Well, you can check it out over at:
http://www.fisixengine.com/demos/demo7.html

It's still a bit slow because of the blob rendering algo. Im using, but hopefully I'll find a better one that will speed it up in the future.

Oz

Wow nice work Ozmic i predict you done it with as3?? Can you use as1, or as2 to pull of this effect?? iv got flash 9 preview but i dont use it lol mostly because i dont want to learn it yet.

Well, technically you code code it up in as1/2 and it would work, the only question in would it run at a decent speed..

Right now, even with as3, the app lags a bit at times so I don't think anything less would hold up, although my algorithm could definitely be made more efficient

I want to start as3 but the language seams way diffrent to as 1 or 2 to me.
Do you make ur as 3 scripts up your self and do you know of any good game as3 tutorials??

I started using as3.0 after having used Java for a good while, so learning it wasnt too much of a leap. But even so, AS3.0 isnt THAT much different from as2.

In any case, the best resource I've found so far is the as2-as3 migration document (http://livedocs.macromedia.com/flex/...migration.html)

Maybe go over this article to get familiar with the language, and of course, check out the AS3 forum over here and its resources sticky

Very cool ozmic, wil you be using in anything?

Well right now I don't have much time to make a real game using this, but perhaps my next project will feature some fun jello creatures

Is the example still up? Nothing happens when it loads.

Internet Explorer 6
Flash Player 8

no flash 9 yet?
I used as3 so player 9 is required

Have not had a need to download player 9 ... until now.

Blobs look great! So do your other demos.

What gives the blobs the circular look at the max boundary points? ie the red blob looks like it could be built from three particles that have a larger than unit radius?

Does your engine handle stacking? Would be curious to know what method you used considering the major performance issues you would need to address (ie the Flash player and AS3). It's hard enough squeezing out frames in C++ with all system resources at ones disposal.

Stacking is actually very stable. That's mostly because I use the verlet method. i'll try to put together a demo to show that off

I have read that a verlet integrator is particularly stable but you could implement resting contact in a number of ways. I was wondering what method you use, to solve resting contact ... particularly multiple simultaneous
contact events? Is your contact resolution coupled to some form of CCD like SAT?

I have been playing with a engine built from the ideas of Mirtich & Canny and others like Guendelman who use an impulse based approach to contact resolution. I think that a using a impulse response for colliding contact and a constraint based method for resting contact could provide a perfect combination. I don't know much at all about a constraint based approach to contact resolution.

Any advice ... mistakes not to make ? I find that Flash presents it's own set of problems due to it's limited processing power and a lot of time can be wasted implementing concepts which will bring the Flash player to it's knees in micro-seconds.

thanks in advance

Actually I don't do anything special for resting contact. With fisix I decided to take a less accurate approach, simply projecting the objects out of the way, which is why it's a lot easier to solve collisions and it still looks good. The only problem with stacking is that if there are too many objects on top of each other they may penetrate, which can be fixed by raising the iterations of that part of the simulation

Your results had me convinced that you must have been using a more complex method.

So your just iterating over your contact list with piece wise resolution in a chronological order ... no special propagation method ?

Are you just projecting along the collision normal by the minimum translation distance ... or do you add some extra distance into the projection?

I know that the number of iterations needed can be very dependant on a particular simulation configuration ... but as a "ball park" figure how many iterations of your collision solver does it take to stack 3 - 4 boxes and what sort of integration step size would you use?

How many stacked boxes would it take to halt the simulation using your current method?

I find this area of multiple simultaneous contact resolution just brings Flash to it's knees ... and am still searching for some insight in how to deal with the need to increase the iterations of the solver. Have you come accross anything of interest that produces plausible results.

Hey Shipstern,

To answer your questions:

Quote:

---

So your just iterating over your contact list with piece wise resolution in a chronological order ... no special propagation method ?

---

Nope, just plain old iterating through the list

Quote:

---

Are you just projecting along the collision normal by the minimum translation distance ... or do you add some extra distance into the projection?

---

minimum distance

looks like you're pretty interested in this, so I put together some videos, take a look:

5 iterations
10 objects stacked
20 objects stacked

25 iterations
10 objects stacked
20 objects stacked

As ou can see, more iterations = more stability, but less fps

hope that helped,
Oz

Hey thanks for putting together the vids. A picture says a thousand words. Either way it's still a very usable frame rate. Congrats for getting it all together and producing the fisix engine !!

I found the frame rate encouraging ... there is hope yet for real-time physics in Flash.

What sort of frame rate would you expect to get with the same configurations with AS1 & AS2 ... or would it just not be viable? Curious as I am still using old AS1 and it may give me an insight into performance gains I could expect should I ever get a more recent version of Flash.

Are you exploiting spatial or frame coherence ? If not it is a good place to get some nice performance boosts that are not reliant on coming up with even more optimized code. Don't know what you are doing with your collision detection but keeping track of closest features and exploiting frame to frame coherence with a list of estimated TOI's can give good gains and it does not require any earth shattering code.

I am sure you have most probably read everything on this site ... but if not ... there are some interesting topics/concepts covered.

http://www.continuousphysics.com/Bul...pBB2/index.php


Thanks again for putting together the vids.

shipstern

Hey Shipstern,

No prob. ;) glad the vids were useful

After reading much about collision detection schemes I've found there are a few good choices, but right now I'm using only one though I'm planning on adding more options in the future.

I've found sweep and prune to work best as it uses frame coherence well and has no memory needs (unlike grid-based methods)

sap has actually made the simulation a lot faster, so without it I don't know how much the frame-rate will drop, but it'd be significant

On AS1/2 you can probably expect the fps to be as much as 10 times lower, but it really depends. hey man, AS2 is almost a joke for anything mathematical, it's time to move to AS3 ;)

ps - thanx for the good link, I've actually been on that site before but never really checked it out thoroughly

This is going to sound really dumb ... maybe ... but is AS3 still in a development stage? For some reason I thought it was ... not keeping up on these things enough!!

If I want to compile AS3 I will have to fork out for the latest version of Flash ... won't I?

I program mostly in C / C++, I really enjoy Flash because it's fun, simple and fast to visualize concepts and ideas. I think it's also a challenge of sorts to squeeze things out of Flash, get your hands dirty and program.

With C / C++ it's all been done all the codes boxed and packaged ... and I know I am not a good enough programmer to do better. It's just too tempting to download the open-source code when after 6 weeks I still can't crack it and I am realizing when it comes to the crunch I am no David Baraff.

With Flash you can still come up with stuff that has not been done in Flash and most of the time that open-source code needs to be totally re-thought for Flash, it makes it more interesting ... I think.

10X faster sound pretty awsome !!!

AS3 has been out of development for a while now, here's how you can start using it:

- get flash 9 alpha which is simply the flash 8 authoring with AS3 compiling
- or get flex builder 2, which is built on eclipse and manages AS3 code well (my preferred choice)
- another choice is using flashdevelop as it is free, unlike flex builder
- you can also just get the command-line compiler and compile away like in the old c days :)

Hey Oz,
Thanks for the info on AS3. It's been a while since I looked at open source Flash etc. so I just had a look around FlashDevelop and Adobe site info on AS3 Flash 9 alpha.

I just have another quick question ... hope you don't mind ... figure it will save me a heap of time.

Will I need ANY runtime libraries etc that would usually come with the Flash IDE or are there alternatives. I noticed that to use Flash 9 alpha you need Flash 8 prof version.

Just seems like I would need more than just a compiler and did not see any alternatives available ... but maybe I just don't know where they are.

Thank again

Shipstern

if you download and install flash 9 alpha, everything you need is included
same with flex builder and flashdevelop

if you want to go the command-line route, there may be a few swc files to include, but I think I've seen some tutorials written about that in adobe live docs

overall it should be very simple to get your as3 code compiled


here's an online as3 compiler.. pretty cool:
http://try.flex.org/index.cfm