CODEWITHSUNDEEP
javaopengllwjgl
maniel34
maniel34

Reputation: 187

FloatBuffer taking up more and more RAM

When I use FloatBuffers, they take up more and more memory over time. I use a float buffer to put a transormation matrix in it, and then I upload it to the shader. So I call the method, which creates the float buffer, every frame.

private static FloatBuffer matrixBuffer;

public void uploadUniformMatrix4f(String name, Matrix4f matrix4f)
{
    // Create a float buffer and put the matrix in it
    matrixBuffer = BufferUtils.createFloatBuffer(16);
    matrix4f.get(matrixBuffer);


    // Upload the float buffer as a matrix
    GL20.glUniformMatrix4fv(getUniformLocation(name), false, matrixBuffer);
}

I already tried memFree(matrixBuffer); and matrixBuffer.clear(); but nothing worked.

The first four slots are getting higher and higher

Upvotes: 2

Views: 360

Answers (1)

Victor Gubin
Victor Gubin

Reputation: 2937

BufferUtils.createFloatBuffer(16) creates new buffer each time your are calling it, outside of JVM heap. I.e. java garbage collector only have reference to the ByteBuffer object, and not for it's data unlike an float[] data = new float[16]; so GC will clear native memory at time it need to clear JVM heap.

BufferUtils approach is not recommended by lwjgl.

LWJGL versions before 3 relied exclusively on allocateDirect(), via the org.lwjgl.BufferUtils class. This class is still there in 3, for backwards-compatibility, but its use is highly discouraged. The reason is simple, allocateDirect() is horrible:

You can use following approach to transfer OpenGL matrices effectively. 

import static org.lwjgl.opengl.GL20.glUniformMatrix4fv;
......
try (MemoryStack stack = MemoryStack.stackPush()) {

   final FloatBuffer matrixBuffer = stack.mallocFloat(16);
   // if you have more matrices like mvp, model, view, projection, normal
   // you don't have to create new memory stack for them
   // simply call stack.mallocFloat(16) as many times as you need  
   matrix4f.get(matrixBuffer);

   glUniformMatrix4fv(getUniformLocation(name), false, matrixBuffer);
   // call OpenGL shader program code here. 
   // When this try block ends - matrixBuffer native memory will be invalid. And OpenGL will crash when accessing the uniform memory already freed   
}

Upvotes: 2

Related Questions