Re: [gl4java-usergroup] newbie question: GLCanvas

[Kenneth B. R. <kbr...@al...>]

> I read somewhere deriving from GLCanvas was the "old way" of doing things.
> Is there a "new way", and if so, which of the two gives the best
> performance?

The new mechanism is a listener mechanism which avoids
subclassing. It offers higher performance, especially for
GLAnimCanvas, because in certain circumstances OpenGL context
switches can be avoided. (The details of this are hidden from the
programmer.)

To use this new mechanism:

  - Create a gl4java.GLCapabilities object indicating the use of
    RGBA mode, etc.
  - Pass this capabilities object to
    gl4java.drawable.GLDrawableFactory.getFactory().
      create[type]
  - Call addGLEventListener on the resulting GLDrawable to set up
    your listener, which should implement at least init() and
    display().

For highest performance:

  - Call GLDrawableFactory.createGLAnimCanvas()
  - Call setUseRepaint(false), setUseFpsSleep(false), and
    setUseYield(false) on the resulting GLAnimCanvas

Note on GLJPanel: the implementation currently renders to an
off-screen buffer and copies the pixels to the lightweight
component. This leads to slowdowns over GLAnimCanvas because the
rendering is not hardware-accelerated and because of the copying
overhead. It is possible that the implementation could be changed
to walk the lightweight component hierarchy to find the bounds of
the GLJPanel within its parent heavyweight component and clip the
OpenGL rendering region to fit (if the GLJPanel itself isn't
overlapped by another lightweight component), which should speed
it up significantly, though this kind of code is tricky to write.