AS3 Animation Base Library Go Performance Tests

Moses has posted some nicely presented info on Go performance.

So far it looks almost neck and neck with TweenLite in drop dead sprints for fastest performing exhaustive tween kung fu-ing.

Although these are not really for competition it is to mainly show patterns for design for purposes that you need. Where this is a more apples to apples comparison (Go vs TweenLite) as the other kits have other overhead such as filters, utilities, even pathing for AnimationPackage and Tweener. TweenLite solved this by separating out into TweenFilterLite and just making TweenLite for animation (and keeping file size extremely small, virtual machine advantages). There is just no excuse for the F9 Tween class though, what the…

It is one thing to build, another to share, another to present information in a very consumable way and then another to make that whole presentation look really good. Moses, like polygonal labs, throws down some nice demos and information posts.

Go was late to the AS3 Animation kit game after pwning with FuseKit in the AS2 age (especially the creative agency love), I think it was the right time as it was released the TweenLite and TweenBencher performance testing utilities put a focus on performance to see just how many more cycles we could get out of AS3 from an animation kit. When building your own animation kits or contributing to one, these observations from various aims helps in the code design.

I still use Tweener in most production work, and TweenLite when I need really small assets if there are going to be many of them. But, I have started to use Go in a kit I am building that I hope to share more in the future, and used it in a small game. But by the speed of the tests and my own experiments it is pretty clear to see that both TweenLite and Go would be excellent base layers to animation kit architecture and Tweener is a bit more on top of that with the filters and bezier tools that it is really a more complete package with less work to do as your animation gets more complex (colors, saturation, bezier, etc). But if you were looking to build your own animation kit or for micro assets a base like Go or a base kit like TweenLite is the way to go). Some notes from Moses’s tests show that performance and sync are also what starts to fall apart as performance critical mass is reached. Go and Tweener held sync the best.

Side notes:

The TweenLite system was highly performative on all three measures. That system also features a very small filesize footprint, making it a clear choice for banners or other filesize-restrictive projects.

Go & Tweener were the only systems here that synced their animations – others ran out of sync to different degrees which yields less visually favorable results. Actually, it looks kind of neat in the tests! But you don’t want out-of-sync animation in your real-life projects. This effect can be clearly seen using the open-source TweenBencher utility, included in the Go package.

  • http://bog.bouguerra.org bougie

    Can you give me the piece of code that you used to test the tweens? I am going to test mine and see where it stands.

  • http://drawlogic.com/ drawk

    Moses used TweenBencher. http://code.google.com/p/goasap/wiki/TweenBencher

    Also, in both benchmarks TweenLite and Go the authors of the kits their code came out the fastests in both cases. But they do offer the downloads and you can try them for yourself. The kits are all good in different ways and some perform better than others in certain profiles. So benchmarks are very important for relative measures and exhaustive stress testing, but that is one element to design. Simplicity, ease of use, completeness, purpose, whether it is for one big file ro many many assets etc. There are lots of factors to deciding what kit to use for what. I have some notes on this and am going to post soon regarding this based on about 1.5 years of work with as3 so far professionally and personal contract work and hobby. I have found I like to use certain kits in certain situations. Speed is very important though, most flash is client side so it is not as much a stress on the server but the more cycles you can save to focus on your main elements is always money.

  • http://drawlogic.com/ drawk

    Here is teh Jack from GreenSock test (author of TweenLite)

    http://blog.greensock.com/tweening-speed-test/

    I validated both of them and they are similar relative results on my machines. Try wiring your kits into these test units and see what you get for your own animation kit.

  • Pingback: TweenBenchMark on DrawLogic: Labs@flexcoders