<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML//EN"> <HTML> <HEAD> <TITLE>Jogl - User's Guide</TITLE> </HEAD> <BODY> <H1>Jogl - User's Guide</H1> <P> <UL> <LI> Overview <LI> Creating a GLDrawable <LI> Writing a GLEventListener <LI> Using the Composable Pipeline <LI> Multithreading Issues <LI> Pbuffers <LI> Platform notes <UL> <LI> All Platforms <LI> Windows <LI> Solaris, Linux (X11 platforms) <LI> Macintosh OS X </UL> </UL> <H2> Overview </H2> <P> Jogl is a Java programming language binding for the OpenGL 3D graphics API. It supports integration with the Java platform's AWT and Swing widget sets while providing a minimal and easy-to-use API that handles many of the issues associated with building multithreaded OpenGL applications. Jogl provides access to the latest OpenGL routines (OpenGL 1.4 with vendor extensions) as well as platform-independent access to hardware-accelerated offscreen rendering ("pbuffers"). Jogl also provides some of the most popular features introduced by other Java bindings for OpenGL like GL4Java, LWJGL and Magician, including a composable pipeline model which can provide faster debugging for Java-based OpenGL applications than the analogous C program. </P> <P> Jogl was designed for the most recent version of the Java platform and for this reason supports only J2SE 1.4 and later. It also only supports truecolor (15 bits per pixel and higher) rendering; it does not support color-indexed modes. Certain areas of the public APIs are more restrictive than in other bindings; for example, the GLCanvas and GLJPanel classes are final, unlike in GL4Java, and the GLContext class is no longer exposed in the public API. These changes have been made to keep the public API simple and because most of the programming errors that have been seen with earlier Java/OpenGL interfaces, in particular GL4Java, have been related to subclassing the OpenGL widget classes and performing manual OpenGL context management. Several complex and leading-edge OpenGL demonstrations have been successfully ported from C/C++ to Jogl without needing direct access to any of these APIs. However, all of these classes and concepts are accessible at the Java programming language level in implementation packages, and in fact the Jogl binding is itself written almost completely in the Java programming language. There are roughly 150 lines of handwritten C code in the entire Jogl source base (100 of which work around bugs in older OpenGL drivers on Windows); the rest of the native code is autogenerated during the build process by a new tool called GlueGen, the source code of which is in the Jogl source tree. Documentation for GlueGen is forthcoming. </P> <H2> Creating a GLDrawable </H2> <P> Jogl provides two basic widgets into which OpenGL rendering can be performed. The GLCanvas is a heavyweight AWT widget which supports hardware acceleration and which is intended to be the primary widget used by applications. The GLJPanel is a fully Swing-compatible lightweight widget which currently does not support hardware acceleration but which is intended to provide 100% correct Swing integration in the rare circumstances where a GLCanvas can not be used. See <a href = "http://java.sun.com/products/jfc/tsc/articles/mixing/">this article</a> on <a href = "http://java.sun.com/products/jfc/tsc/">The Swing Connection</a> for more information about mixing lightweight and heavyweight widgets. </P> <P> Both the GLCanvas and GLJPanel implement a common interface called GLDrawable so applications can switch between them with minimal code changes. The GLDrawable interface provides <UL> <LI> access to the GL and GLU objects for calling OpenGL routines <LI> the mechanism for registering GLEventListeners for performing OpenGL rendering <LI> a <CODE>display()</CODE> method for forcing OpenGL rendering to be performed <LI> exclusion methods (<CODE>setRenderingThread()</CODE>, <CODE>setNoAutoRedrawMode()</CODE>) for controlling the multithreading behavior of the widget <LI> AWT- and Swing-independent abstractions for getting and setting the size of the widget and adding and removing event listeners <LI> a platform-independent mechanism for creating hardware-accelerated offscreen surfaces (pbuffers) for performing advanced rendering techniques </UL> </P> <P> GLCanvas and GLJPanel instances are created using the factory methods in GLDrawableFactory. These factory methods allow the user to request a certain set of OpenGL parameters in the form of a GLCapabilities object as well as optionally customize the format selection algorithm by specifying a GLCapabilitiesChooser. </P> <P> A GLCapabilities object specifies the OpenGL parameters for a newly-created widget, such as the color, alpha,, z-buffer and accumulation buffer bit depths and whether the widget is double-buffered. The default capabilities are loosely specified but provide for truecolor RGB, a reasonably large depth buffer, double-buffered, with no alpha, stencil, or accumulation buffers. </P> <P> An application can override the default pixel format selection algorithm by providing a GLCapabilitiesChooser to the GLDrawableFactory. The chooseCapabilities method will be called with all of the available pixel formats as an array of GLCapabilities objects; it should return an integer index into this array. The DefaultGLCapabilitiesChooser attempts to provide a better cross-platform selection algorithm than the WGL and GLX pixel format selection algorithms. </P> <H2> Writing a GLEventListener </H2> <P> Applications implement the GLEventListener interface to perform OpenGL drawing. When the methods of the GLEventListener are called, the underlying OpenGL context associated with the drawable is already current. The listener fetches the GL object out of the GLDrawable and begins to perform rendering. </P> <P> The <CODE>init()</CODE> method is called upon OpenGL context creation. Any display lists or textures used during the application's normal rendering loop can be safely initialized in <CODE>init()</CODE>. Because the underlying AWT window may be destroyed and recreated while using the same GLCanvas and GLEventListener, the GLEventListener's <CODE>init()</CODE> method may be called more than once during the lifetime of the application. It is the responsibility of the application to understand its sharing of textures and display lists between multiple OpenGL contexts and reinitialize them when the <CODE>init()</CODE> callback is entered if necessary. </P> <P> The <CODE>display()</CODE> method is called to perform per-frame rendering. The <CODE>reshape()</CODE> method is called when the drawable has been resized; the default implementation automatically resizes the OpenGL viewport so often it is not necessary to do any work in this method. The <CODE>displayChanged()</CODE> method is designed to allow applications to support on-the-fly screen mode switching, but support for this is not yet implemented so the body of this method should remain empty. </P> <P> It is strongly recommended that applications always refetch the GL and GLU objects out of the GLDrawable upon each call to the <CODE>init()</CODE>, <CODE>display()</CODE> and <CODE>reshape()</CODE> methods and pass the GL object down on the stack to any drawing routines, as opposed to storing the GL in a field and referencing it from there. The reason is that multithreading issues inherent to the AWT toolkit make it difficult to reason about which threads certain operations are occurring on, and if the GL object is stored in a field it is unfortunately too easy to accidentally make OpenGL calls from a thread that does not have a current context. This will usually cause the application to crash. For more information please see the section on multithreading. </P> <H2> Using the Composable Pipeline </H2> <P> Jogl supports the "composable pipeline" paradigm introduced by the Magician Java binding for OpenGL. The DebugGL pipeline calls <CODE>glGetError</CODE> after each OpenGL call, reporting any errors found. It can greatly speed up development time because of its fine-grained error checking as opposed to the manual error checking usually required in OpenGL programs written in C. The TraceGL prints logging information upon each OpenGL call and is helpful when an application crash makes it difficult to see where the error occurred. </P> <P> To use these pipelines, call <CODE>GLDrawable.setGL</CODE> at the beginning of the <CODE>init</CODE> method in your GLEventListener. For example, <PRE> class MyListener implements GLEventListener { public void init(GLDrawable drawable) { drawable.setGL(new DebugGL(drawable.getGL())); // ... } // ... } </PRE> </P> <H2> Multithreading Issues </H2> <P> Jogl was designed to interoperate with the AWT, an inherently multithreaded GUI toolkit. OpenGL, in contrast, was originally designed in single-threaded C programming environments. For this reason Jogl provides a framework in which it is possible to write correct multithreaded OpenGL applications using the GLEventListener paradigm. </P> <P> If an application written using Jogl interacts in any way with the mouse or keyboard, the AWT is processing these events and the multithreaded aspects of the program must be considered. </P> <P> OpenGL applications usually behave in one of two ways: either they repaint only on demand, for example when mouse input comes in, or they repaint continually, regardless of whether user input is coming in. In the repaint-on-demand model, the application can merely call <CODE>GLDrawable.display()</CODE> manually at the end of the mouse or keyboard listener to cause repainting to be done. Alternatively if the application knows the concrete type of the GLDrawable it can call repaint() to have the painting scheduled for a later time. </P> <P> In the continuous repaint model, the application typically has a main loop which is calling <CODE>GLDrawable.display()</CODE> repeatedly, or is using the Animator class, which does this internally. In both of these cases the OpenGL rendering will be done on this thread rather than the internal AWT event queue thread which dispatches mouse and keyboard events. </P> <P> Both of these models (repaint-on-demand and repaint continually) still require the user to think about which thread keyboard and mouse events are coming in on, and which thread is performing the OpenGL rendering. OpenGL rendering <B>may not</B> occur directly inside the mouse or keyboard handlers, because the OpenGL context for the drawable is not current at this point (hence the warning about storing a GL object in a field, where it can be fetched and accidentally used by another thread). However, a mouse or keyboard listener may invoke <CODE>GLDrawable.display()</CODE>. </P> <P> It is generally recommended that applications perform as little work as possible inside their mouse and keyboard handlers to keep the GUI responsive. However, since OpenGL commands can not be run from directly within the mouse or keyboard event listener, the best practice is to store off state when the listener is entered and retrieve this state during the next call to <CODE>GLEventListener.display()</CODE>. </P> <P> Furthermore, it is recommended that if there are long computational sequences in the GLEventListener's <CODE>display</CODE> method which reference variables which may be being simultaneously modified by the AWT thread (mouse and keyboard listeners) that copies of these variables be made upon entry to <CODE>display</CODE> and these copies be referenced throughout display() and the methods it calls. This will prevent the values from changing while the OpenGL rendering is being performed. Errors of this kind show up in many ways, including certain kinds of flickering of the rendered image as certain pieces of objects are rendered in one place and other pieces are rendered elsewhere in the scene. Restructuring the display() method as described has solved all instances of this kind of error that have been seen with Jogl to date. </P> <P> Prior to JOGL 1.1 b10, the JOGL library attempted to give applications strict control over which thread or threads performed OpenGL rendering. The <CODE>setRenderingThread()</CODE>, <CODE>setNoAutoRedrawMode()</CODE> and <CODE>display()</CODE> APIs were originally designed to allow the application to create its own animation thread and avoid OpenGL context switching on platforms that supported it. Unfortunately, serious stability issues caused by multithreading bugs in either vendors' OpenGL drivers or in the Java platform implementation have arisen on three of JOGL's major supported platforms: Windows, Linux and Mac OS X. A detailed description of these issues can be found on <a href="http://www.javagaming.org/cgi-bin/JGNetForums/YaBB.cgi?board=jogl;action=display;num=1109286717">this thread</a> in the <a href="http://www.javagaming.org/cgi-bin/JGNetForums/YaBB.cgi?board=jogl">JOGL forums</a>. In order to address these bugs, the threading model in JOGL 1.1 b10 and later has changed. </P> <P> All GLEventListener callbacks and other internal OpenGL context management are now performed on one thread: the AWT event queue thread. This is a thread internal to the implementation of the AWT and is always present when the AWT is being used. When the <CODE>GLDrawable.display()</CODE> method is called from user code, it now performs the work synchronously on the AWT event queue thread, even if the calling thread is a different thread. The <CODE>setRenderingThread()</CODE> optimization is now a no-op. The <CODE>setNoAutoRedraw()</CODE> API still works as previously advertised, though now that all work is done on the AWT event queue thread it no longer needs to be used in most cases. (It was previously useful for working around certain kinds of OpenGL driver bugs.) </P> <P> Most JOGL applications will not see a change in behavior from this change in the JOGL implementation. Applications which use thread-local storage or complex multithreading and synchronization may see a change in their control flow requiring code changes. While it is strongly recommended to change such applications to work under the new threading model, the old threading model can be used by specifying the system property <CODE>-Djogl.1thread=auto</CODE> or <CODE>-Djogl.1thread=false</CODE>. The "auto" setting is equivalent to the behavior in 1.1 b09 and before, where on ATI cards the single-threaded mode would be used. The "false' setting is equivalent to disabling the single-threaded mode. "true" is now the default setting. </P> <H2> Pbuffers </H2> <P> Jogl exposes hardware-accelerated offscreen rendering (pbuffers) with a minimal and platform-agnostic API. Several recent demos have been successfully ported from C/C++ to Java using Jogl's pbuffer APIs. However, the pbuffer support in Jogl remains one of the more experimental aspects of the package and the APIs may need to change in the future. </P> <P> To create a pbuffer, create a GLCanvas and (assuming it reports that it can create an offscreen drawable) make a pbuffer using the <CODE>createOffscreenDrawable</CODE> API. Because of the multithreaded nature of the AWT, the pbuffer is actually created lazily. For this reason the application's main loop typically needs to detect when the init() methods of all of the GLEventListeners for all of the offscreen surfaces have been called. See the demonstrations such as the ProceduralTexturePhysics demo for an example of this. </P> <P> Additionally, pbuffers are only created when the parent GLCanvas's display(), init(), or reshape() methods are called; in other words, it may be necessary to manually "prime" the GLCanvas by calling display() on it until it creates all of its requested pbuffers. Again, please see the demonstrations for concrete examples of this. We hope that it may be possible to hide many of these details in the future. </P> <P> A pbuffer is used by calling its display() method. Rendering, as always, occurs while the pbuffer's OpenGL context is current. There are render-to-texture options that can be specified in the GLCapabilities for the pbuffer which can make it easier to operate upon the resulting pixels. These APIs are however extremely experimental and not yet implemented on all platforms. </P> <H2> Platform Notes </H2> <H3> All Platforms </H3> <P> The following issues, among others, are outstanding on all platforms: </P> <UL> <LI> A few remaining stability issues, mostly on older graphics cards. </UL> <H3> Windows </H3> <P> No outstanding issues at this time. </P> <H3> Solaris, Linux (X11 platforms) </H3> <P> No outstanding issues at this time. </P> <H3> Mac OS X </H3> <P> The Mac OS X port of Jogl, in particular the GL interface and its implementation, can be used either with the provided GLCanvas widget or with the Cocoa NSOpenGLView. In order to use it with Cocoa the following steps should be taken: <UL> <LI> Instantiate a <CODE>net.java.games.jogl.impl.macosx.MacOSXGLImpl</CODE> using the public constructor taking no arguments. <LI> Upon the first render of your application, or whenever the available OpenGL routines might have changed (because a window moved from one screen to another) call the publicly-accessible method <CODE>MacOSXGLImpl.resetGLFunctionAvailability()</CODE>. <LI> Only use the GL instance when the OpenGL context from the NSOpenGLView is current. </UL> <B>NOTE:</B> the Cocoa interoperability has not yet been retested since the GLCanvas was implemented. Please report any problems found with using Jogl with an NSOpenGLView. </P> <P> The following issues remain with the Mac OS X port: <UL> <LI> Due to the mechanism by which the Cocoa graphics system selects OpenGL pixel formats, the GLCapabilitiesChooser mechanism can not be implemented on Mac OS X as on other platforms. In the future, the chooser will be used and the capabilities array will contain exactly the requested capablities. The underlying Cocoa pixel format selection algorithm will then run to choose the best-fit visual. Currently the capabilities of the underlying context are hardcoded. <LI> Pbuffers are supported in Jogl on Mac OS X 10.3 and later. </UL> </P> </BODY> </HTML>