Wednesday, October 12, 2016

Godot (GDScript) Memory Management

I've been working with the cocos2d-x engine for a while and have gotten used to one of the important features in the engine.. the autorelease().

It's basically a reference counting implementation which automatically deletes any object that is removed from a scene. Unless ofcourse you called a retain on it. So when i was working on DotDot a simple 2D game, i created a crosshair animation everytime the user got a perfect alignment and removed the object once the animation was done. It worked great and for a while i assumed that everything was fine.

I was showing off the game to people on my android device when i started noticing slowdowns once in a while.. and as the score got higher it got worse.. surely the engines' fault i assumed.. but then i thought ill' give Godot the benefit of doubt and checked the documentation about freeing the memory. That is when i discovered that the objects removed from the scene are not actually removed from memory. Basically every time i did a perfect align in the game i was leaking memory. I used the inbuilt profiler to try and figure out which object was getting left out in the memory and it was the cross hair object.

So how do we get around this problem?.. Ideally we should be maintaining a pool of the objects and changing position / re animating the object.. but for the game it wasn't that important. I chose the method of creating / destroying the object instead and for this to work i had to use "queue_free()" call in the "_exit_tree()" which is called when the node is removed from tree.

Hope this helps anyone trying to do something similar, this was definitely a big miss and could become a problem if you miss it :)

Sunday, October 02, 2016

Pixel perfect mouse click on Images in Android Canvas

I was trying to build a sample for some image manipulation tool. The requirement was that the user should be able to pick an image that he pasted on the screen, move it around, scale and rotate it.

The way we would do this using OpenGL is create a depth buffer use different depth value on each of the Quads' used to render the transparent image. So when you do a readPixel from the depth buffer you will get the z value of the sprite if it was under the touch. This is how we wold distinguish between two overlapping transparent images.

The tracking of z value for different Quads is done automatically in OpenGL when you bind the depth buffer to the current frame buffer and start drawing stuff. Now this i know and have used for a while, but how does this work in the Android Canvas drawing world?

Short answer: in the same way
Long answer: 
These are the steps you should follow to get it done:
  • You have to create a separate Bitmap of the same size as your drawing area
  • Set the Bitmap.Config parameter to Bitmap.Config.A_8
  • Bind a canvas to the buffer. 
  • on Touch event, set the clipping rect of the backbuffer canvas to a rect around the touch location.
  • Read the pixel value from the back buffer bitmap. 
  • Compare the value of pixel to see if it matches any of the sprites. If it does then you have a touch on image else its not.
Following is the function for detecting the pixel value from the backbuffer based on touch location.


This should help you get ahead if you are stuck with the same problem :)