NIO ByteBuffer

When it comes to OpenGL with Java, we use NIO ByteBuffer for off-heap memory a lot.

Here are some tips:

limit - position = remaining. OpenGL functions (LWJGL ones) read from the current position up to limit, i.e., the remaining() bytes.
So, set position = offset, limit = byteCount + offset before passing to GL. Be cautious when using views — intBuffer.limit() is in ints, not bytes!
Use direct buffers (ByteBuffer.allocateDirect) for OpenGL. Only direct buffers are compatible with native GPU memory access.
Never allocate byte buffers in hotpaths (slow!)
Therefore, only allocate once, typically in an init/setup method. Reusing buffers reduces GC pressure and avoids repeated native allocations.
Don't forget to set order to nativeOrder (aka CPU endianness)
ByteBuffer.asFloatBuffer()/ByteBuffer.<any other interpretation> points to the same underlying memory, but returns a new wrapper. These views have their own position/limit/capacity measured in elements (e.g., ints), not bytes. Adding contents to alternative views doesn't grow the original ByteBuffer (i.e. grow pos etc.)
ByteBuffer.duplicate() doesn't copy endianness but forces big endian. Don't forget to set order again.

Contributors

Last edited about 1 hour ago

View full history