Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Tween Question (Part Deux..)

edited February 2013 in Questions Posts: 398

Experimenting a bit more with the tween function and was wondering how this works 'under the hood'.

a) How does Codea keep track of 'active tweens' during each update cycle? Is there an internal table of 'TweenIds' that are exposed and can be queried or is this something we have to keep track of ourselves?

b) If I stop a Tween animation before it completes with 'tween.stop' (e.g. a tweened sprite being hit and destroyed in a game) does this automatically get garbage collected at somepoint or do we have to force this to occur by setting the TweenId to 'nil'? I'm trying to work out a best practice method for exiting a tween gracefully and cleaning up after itself, else I fear this simply keeps it in memory.

c) In respect to speed, has anyone done any benchmarks regarding a large number of multiple tweens running simultaneously?

Any thoughts? :-D

Tagged:

Comments

  • SimeonSimeon Admin Mod
    Posts: 5,778

    a) Codea keeps a table of active tweens which it updates internally. I believe the table is not exposed.

    b) If you stop a tween — and no longer keep a reference to its ID in your code — it will be garbage collected.

    c) Speed would be interesting to see. The tween library is entirely Lua based (we will be putting our tween.lua on Github so people can look at it and potentially make it better).

  • Hi @Simeon - thanks for this. :-)

    I'm in the process of doing some speed tests with multiple tweens and seeing what the performance is like and will report back. Would be interesting to see the code for this at some point.

    From its functionality, i suspect it's likely to have to cover a lot of ground and might not be necessarily optimised for speed - but let's see.

  • dave1707dave1707 Mod
    Posts: 9,978

    I did some FPS tests with multiple Tweens running for 10 seconds from the bottom of the screen to the top. I did 2 runs on each tween count. Here's the results I got.


    # Tweens 1st run 2nd run 50 59.6375 59.9082 FPS 100 53.0186 52.5563 150 39.8047 40.5957 200 33.2133 32.7913 250 27.6686 28.0826 300 23.8152 23.6842 500 15.7165 15.4772 1000 8.6732 8.8125
  • Thanks @dave1707 , were these with Sprites or Meshes? Did a similar quick test last night (with meshes) and the performance was much better than these - around 60fps for a few hundred. I'll re-run it again and post my timings.

  • SimeonSimeon Admin Mod
    edited February 2013 Posts: 5,778

    .@dave1707 I did the test without any graphics — just N tweens running for 10 seconds, each one interpolating two values [x,y]. These were my results:

    easing = tween.easing.linear
    
    # Tweens                FPS
       50                    60
      100                    60
      500                    52
     1000                    28.5
    
    easing = tween.easing.quadInOut
    
    # Tweens                FPS
       50                    60
      100                    60
      500                    46.5
     1000                    26
    
    easing = tween.easing.elasticInOut
    
    # Tweens                FPS
       50                    60
      100                    60
      500                    36.5
     1000                    19.6
    

    As you can see, the complexity of the easing function plays a big part.

    Also note that interpolating a single value is faster than two values — 1000 elasticInOut tweens interpolating a single value gives me ~30 FPS on iPad 3.

  • @Simeon - I guess I can live without any of the easing functions for a lot of stuff, I guess the tween.lua lib does is a 'simple drop through' if it doesn't detect an easing parameter. Whats the performance like on your test without any easing? I guess thats the best case scenario?

  • SimeonSimeon Admin Mod
    edited February 2013 Posts: 5,778

    .@andymac3d no easing is the same as linear — it's the simplest way for the tween library to interpolate between two values.

    Most of the time you'd want something like quadInOut, which gives you smooth acceleration and deceleration. Elastic and bounce are good for making things pop and jump.

    I wouldn't hold back on using them, spawning 1000, or even 100 simultaneous tweens seems like a very unusual thing to do. Even 10 is quite a lot to instantiate simultaneously.

    If you just use them where you need they will not have any performance penalty.

  • Yes, I guess 1000s are overkill! ;-). Just curious how far to push the hardware really when using tweens - I agree that significantly less will be more than sufficient for most things!

    I am writing something that does use a fair few (>100) simultaneous animation events so I'm keen to see how this performs in practice in comparison to my..ahem.. 'less sophisticated' animation lib.

    I think I'll be using tween from now on.. ;-)

  • dave1707dave1707 Mod
    Posts: 9,978

    Here's the program I used for my timings. The Tweens are mostly default values and the graphics simple. I changed this from my original a little to try and speed it up.


    displayMode(FULLSCREEN) supportedOrientations(PORTRAIT) function setup()     tab={}     average=false     tot=0     count=0     for z=1,50 do    -- change # of tweens, 50, 100, etc.         table.insert(tab,{x=5*z,y=100})         tween(10,tab[z],{x=5*z,y=900})     end end function draw()     background(40, 40, 50)     fill(255)     for z=1,#tab do         ellipse(tab[z].x,tab[z].y,5)     end     if average then    -- show average FPS         text(#tab.." Tweens          Average FPS  "..tot/count,300,1000)     elseif #tab>0 and tab[1].y==900 then         average=true     else         count = count + 1         tot=tot+1/DeltaTime     end end
Sign In or Register to comment.