I have a dog, in scene-1, the dog will be set to texture1 which named "dog".

In scene-2,  i first removeAndUnload("dog"), then new texture2 which is different from texture1 but has the same name "dog", then set the texture2 to the dog.

But the dog in scene-2 looks like with another texture which is not texture2.

Codes:

//scene1
texture1 = new texture(in);
textManager.addTexture("dog", texture1);
dog.setTexture("dog");

//do something
...

//switch to scene2
removeAndUnload("dog");

//in scene2
//add some other texture first
textureManager.addTexture("some other textures")

//new texture2 with the same name "dog"
texture2 = new texture(in2);
textureManager.addTexture("dog", texture2);

dog.setTexture("dog");

// the dog seems not to be set to  the texture2, but one of "some other textures"
I guess that was caused by some caching mechanism?

And you say:

obj=null;
obj=new Object();
insted of

obj=new Object();
I am a new  Javaer, could you explain more? Does the previous code make the GC happier(So, the memory of the obj will be freed fastter)?

I want to use the Virtualizer class of jPCT-AE to resolve OOM, and i have read the doc, but i want to know a little more, thanks.

The process of my app is like this:

Load all models(.ser and .bones) ->
entry Scene1 -> Load all textures which are used in scene1 only -> some logic\event\draw ->
switch to Scene2 -> removeAndUnload all textures(Config.unloadImmediately = true) ->
entry Scene2 -> Load all textures which are used in scene2 only -> repeat like scene1 ... ...
-> switch to SceneN ... ...

And the OOM usually happened in the stage of "entry new scene -> load textures", my question:

1, Does the process of loading texture need more memory（peak) than the texture itself? For example:  new Texture("1024X1024.jpg"), the texture will consume 1024x1024x4 bytes of memory finally, but does the process of "new Texture(like io、decode-jpg and so on)" need more memory than 4MB?     

2, Does TextureManager.removeAndUnload () guarantee freeing memory immediately ?

3, Does the follow memory-route right?

new Texture(1024X1024) -> 4MB

set Virtualizer -> set context -> virtualize the texture -> swap the texture from vm to disk(flash) -> 0MB

Frambuffer.draw() -> upload the texture to GPU -> I don't know what happen with the virtualized-texture in this stage, help please.

removeAndUnload the virtualized-texture -> what happened?

Does the virtualized-texture need be swapped back(from disk to vm) in any situation? When? Does the process of virtualizing need more memory(peak) than the texture itself? How much time does the process of virtualizing consume typically?

4, Seems there are enough memory in every scene of my app(all of them can run fine alone),  but OOM happened in the scene switching, any suggestions? Many thanks! I have reduced my texture to two jpg(512X512) per scene, and it's hard to reduce more. And i also enabled Texture.4bpp, disabled Texture.Mipmapping.

My codes:

   
    Object3D mDog = xxxxxx; //Loader init
    Object3D obj =  Object3D.createDummyObj();
    obj.addChild(mDog);
    obj.addChild(otherObjs);
    obj.rotateY(1);

   if(mDog.getRotationMatrix().isIdentity()){
      log.d("xx", "true");
  };

   if(mDog.getTranslationMatrix().isIdentity()){
      log.d("xx", "true2");
  };


The result is "true"、"true2".

I want to reset "mDog" to its original position after "obj.rotateY(1)",  and the code "mDog.clearRotation()" is useless. Doesn't the rotation of the parent-Object3D affect the RotationMatrix of children-Object3D?

I will read it. But seems i can't access this address.(http://mobilegameengines.com/interviews_with_mobile_game_engine_developers)

I don't know whether the server can't be access now or it is blocked by China_Greate_Fire_Wall.

Yes, there is only one copy of mesh in memory, but i need to load the same mesh twice, which means more "waiting time" for users.

But i think this is not a serious problem, i can handle this by someway, thanks!

I have an animated-dog, it has many skin-animation-frames(about 2000+, sampled by 24frame/s).

If i load all of them in one time, the left memory will be not enough. So, i want to load 1 - 1000 frames in scene1, and when the world switch to scene2, i will release 1-1000 frames, and load 1000-2000 frames.

How could i do this?