QOpenGLWidget Class

Detailed Description

QOpenGLWidget provides functionality for displaying OpenGL graphics integrated into a Qt application. It is very simple to use: Make your class inherit from it and use the subclass like any other QWidget, except that you have the choice between using QPainter and standard OpenGL rendering commands.

QOpenGLWidget provides three convenient virtual functions that you can reimplement in your subclass to perform the typical OpenGL tasks:

• paintGL() - Renders the OpenGL scene. Gets called whenever the widget needs to be updated.
• resizeGL() - Sets up the OpenGL viewport, projection, etc. Gets called whenever the widget has been resized (and also when it is shown for the first time because all newly created widgets get a resize event automatically).
• initializeGL() - Sets up the OpenGL resources and state. Gets called once before the first time resizeGL() or paintGL() is called.

If you need to trigger a repaint from places other than paintGL() (a typical example is when using timers to animate scenes), you should call the widget's update() function to schedule an update.

Your widget's OpenGL rendering context is made current when paintGL(), resizeGL(), or initializeGL() is called. If you need to call the standard OpenGL API functions from other places (e.g. in your widget's constructor or in your own paint functions), you must call makeCurrent() first.

All rendering happens into an OpenGL framebuffer object. makeCurrent() ensure that it is bound in the context. Keep this in mind when creating and binding additional framebuffer objects in the rendering code in paintGL(). Never re-bind the framebuffer with ID 0. Instead, call defaultFramebufferObject() to get the ID that should be bound.

QOpenGLWidget allows using different OpenGL versions and profiles when the platform supports it. Just set the requested format via setFormat(). Keep in mind however that having multiple QOpenGLWidget instances in the same window requires that they all use the same format, or at least formats that do not make the contexts non-sharable. To overcome this issue, prefer using QSurfaceFormat::setDefaultFormat() instead of setFormat().

Painting Techniques

As described above, subclass QOpenGLWidget to render pure 3D content in the following way:

• Reimplement the initializeGL() and resizeGL() functions to set up the OpenGL state and provide a perspective transformation.
• Reimplement paintGL() to paint the 3D scene, calling only OpenGL functions.

It is also possible to draw 2D graphics onto a QOpenGLWidget subclass using QPainter:

• In paintGL(), instead of issuing OpenGL commands, construct a QPainter object for use on the widget.
• Draw primitives using QPainter's member functions.
• Direct OpenGL commands can still be issued. However, you must make sure these are enclosed by a call to the painter's beginNativePainting() and endNativePainting().

When performing drawing using QPainter only, it is also possible to perform the painting like it is done for ordinary widgets: by reimplementing paintEvent().

• Reimplement the paintEvent() function.
• Construct a QPainter object targeting the widget. Either pass the widget to the constructor or the QPainter::begin() function.
• Draw primitives using QPainter's member functions.
• Painting finishes then the QPainter instance is destroyed. Alternatively, call QPainter::end() explicitly.

OpenGL Function Calls, Headers and QOpenGLFunctions

When making OpenGL function calls, it is strongly recommended to avoid calling the functions directly. Instead, prefer using QOpenGLFunctions (when making portable applications) or the versioned variants (for example, QOpenGLFunctions_3_2_Core and similar, when targeting modern, desktop-only OpenGL). This way the application will work correctly in all Qt build configurations, including the ones that perform dynamic OpenGL implementation loading which means applications are not directly linking to an GL implementation and thus direct function calls are not feasible.

In paintGL() the current context is always accessible by calling QOpenGLContext::currentContext(). From this context an already initialized, ready-to-be-used QOpenGLFunctions instance is retrievable by calling QOpenGLContext::functions(). An alternative to prefixing every GL call is to inherit from QOpenGLFunctions and call QOpenGLFunctions::initializeOpenGLFunctions() in initializeGL().

As for the OpenGL headers, note that in most cases there will be no need to directly include any headers like GL.h. The OpenGL-related Qt headers will include qopengl.h which will in turn include an appropriate header for the system. This might be an OpenGL ES 3.x or 2.0 header, the highest version that is available, or a system-provided gl.h. In addition, a copy of the extension headers (called glext.h on some systems) is provided as part of Qt both for OpenGL and OpenGL ES. These will get included automatically on platforms where feasible. This means that constants and function pointer typedefs from ARB, EXT, OES extensions are automatically available.

Code Examples

To get started, the simplest QOpenGLWidget subclass could look like the following:

class MyGLWidget : public QOpenGLWidget
{
public:
    MyGLWidget(QWidget *parent) : QOpenGLWidget(parent) { }

protected:
    void initializeGL() override
    {
        // Set up the rendering context, load shaders and other resources, etc.:
        QOpenGLFunctions *f = QOpenGLContext::currentContext()->functions();
        f->glClearColor(1.0f, 1.0f, 1.0f, 1.0f);
        ...
    }

    void resizeGL(int w, int h) override
    {
        // Update projection matrix and other size related settings:
        m_projection.setToIdentity();
        m_projection.perspective(45.0f, w / float(h), 0.01f, 100.0f);
        ...
    }

    void paintGL() override
    {
        // Draw the scene:
        QOpenGLFunctions *f = QOpenGLContext::currentContext()->functions();
        f->glClear(GL_COLOR_BUFFER_BIT);
        ...
    }

};

Alternatively, the prefixing of each and every OpenGL call can be avoided by deriving from QOpenGLFunctions instead:

class MyGLWidget : public QOpenGLWidget, protected QOpenGLFunctions
{
    ...
    void initializeGL() override
    {
        initializeOpenGLFunctions();
        glClearColor(...);
        ...
    }
    ...
};

To get a context compatible with a given OpenGL version or profile, or to request depth and stencil buffers, call setFormat():

QOpenGLWidget *widget = new QOpenGLWidget(parent);
QSurfaceFormat format;
format.setDepthBufferSize(24);
format.setStencilBufferSize(8);
format.setVersion(3, 2);
format.setProfile(QSurfaceFormat::CoreProfile);
widget->setFormat(format); // must be called before the widget or its parent window gets shown

With OpenGL 3.0+ contexts, when portability is not important, the versioned QOpenGLFunctions variants give easy access to all the modern OpenGL functions available in a given version:

    ...
    void paintGL() override
    {
        QOpenGLFunctions_3_2_Core *f = QOpenGLContext::currentContext()->versionFunctions<QOpenGLFunctions_3_2_Core>();
        ...
        f->glDrawArraysInstanced(...);
        ...
    }
    ...

As described above, it is simpler and more robust to set the requested format globally so that it applies to all windows and contexts during the lifetime of the application. Below is an example of this:

int main(int argc, char **argv)
{
    QApplication app(argc, argv);

    QSurfaceFormat format;
    format.setDepthBufferSize(24);
    format.setStencilBufferSize(8);
    format.setVersion(3, 2);
    format.setProfile(QSurfaceFormat::CoreProfile);
    QSurfaceFormat::setDefaultFormat(format);

    MyWidget widget;
    widget.show();

    return app.exec();
}

Multisampling

To enable multisampling, set the number of requested samples on the QSurfaceFormat that is passed to setFormat(). On systems that do not support it the request may get ignored.

Multisampling support requires support for multisampled renderbuffers and framebuffer blits. On OpenGL ES 2.0 implementations it is likely that these will not be present. This means that multisampling will not be available. With modern OpenGL versions and OpenGL ES 3.0 and up this is usually not a problem anymore.

Threading

Performing offscreen rendering on worker threads, for example to generate textures that are then used in the GUI/main thread in paintGL(), are supported by exposing the widget's QOpenGLContext so that additional contexts sharing with it can be created on each thread.

Drawing directly to the QOpenGLWidget's framebuffer outside the GUI/main thread is possible by reimplementing paintEvent() to do nothing. The context's thread affinity has to be changed via QObject::moveToThread(). After that, makeCurrent() and doneCurrent() are usable on the worker thread. Be careful to move the context back to the GUI/main thread afterwards.

Triggering a buffer swap just for the QOpenGLWidget is not possible since there is no real, onscreen native surface for it. It is up to the widget stack to manage composition and buffer swaps on the gui thread. When a thread is done updating the framebuffer, call update() on the GUI/main thread to schedule composition.

Extra care has to be taken to avoid using the framebuffer when the GUI/main thread is performing compositing. The signals aboutToCompose() and frameSwapped() will be emitted when the composition is starting and ending. They are emitted on the GUI/main thread. This means that by using a direct connection aboutToCompose() can block the GUI/main thread until the worker thread has finished its rendering. After that, the worker thread must perform no further rendering until the frameSwapped() signal is emitted. If this is not acceptable, the worker thread has to implement a double buffering mechanism. This involves drawing using an alternative render target, that is fully controlled by the thread, e.g. an additional framebuffer object, and blitting to the QOpenGLWidget's framebuffer at a suitable time.

Context Sharing

When multiple QOpenGLWidgets are added as children to the same top-level widget, their contexts will share with each other. This does not apply for QOpenGLWidget instances that belong to different windows.

This means that all QOpenGLWidgets in the same window can access each other's sharable resources, like textures, and there is no need for an extra "global share" context.

To set up sharing between QOpenGLWidget instances belonging to different windows, set the Qt::AA_ShareOpenGLContexts application attribute before instantiating QApplication. This will trigger sharing between all QOpenGLWidget instances without any further steps.

Creating extra QOpenGLContext instances that share resources like textures with the QOpenGLWidget's context is also possible. Simply pass the pointer returned from context() to QOpenGLContext::setShareContext() before calling QOpenGLContext::create(). The resulting context can also be used on a different thread, allowing threaded generation of textures and asynchronous texture uploads.

Note that QOpenGLWidget expects a standard conformant implementation of resource sharing when it comes to the underlying graphics drivers. For example, some drivers, in particular for mobile and embedded hardware, have issues with setting up sharing between an existing context and others that are created later. Some other drivers may behave in unexpected ways when trying to utilize shared resources between different threads.

Resource Initialization and Cleanup

The QOpenGLWidget's associated OpenGL context is guaranteed to be current whenever initializeGL() and paintGL() are invoked. Do not attempt to create OpenGL resources before initializeGL() is called. For example, attempting to compile shaders, initialize vertex buffer objects or upload texture data will fail when done in a subclass's constructor. These operations must be deferred to initializeGL(). Some of Qt's OpenGL helper classes, like QOpenGLBuffer or QOpenGLVertexArrayObject, have a matching deferred behavior: they can be instantiated without a context, but all initialization is deferred until a create(), or similar, call. This means that they can be used as normal (non-pointer) member variables in a QOpenGLWidget subclass, but the create() or similar function can only be called from initializeGL(). Be aware however that not all classes are designed like this. When in doubt, make the member variable a pointer and create and destroy the instance dynamically in initializeGL() and the destructor, respectively.

Releasing the resources also needs the context to be current. Therefore destructors that perform such cleanup are expected to call makeCurrent() before moving on to destroy any OpenGL resources or wrappers. Avoid deferred deletion via deleteLater() or the parenting mechanism of QObject. There is no guarantee the correct context will be current at the time the instance in question is really destroyed.

A typical subclass will therefore often look like the following when it comes to resource initialization and destruction:

class MyGLWidget : public QOpenGLWidget
{
    ...

private:
    QOpenGLVertexArrayObject m_vao;
    QOpenGLBuffer m_vbo;
    QOpenGLShaderProgram *m_program;
    QOpenGLShader *m_shader;
    QOpenGLTexture *m_texture;
};

MyGLWidget::MyGLWidget()
    : m_program(0), m_shader(0), m_texture(0)
{
    // No OpenGL resource initialization is done here.
}

MyGLWidget::~MyGLWidget()
{
    // Make sure the context is current and then explicitly
    // destroy all underlying OpenGL resources.
    makeCurrent();

    delete m_texture;
    delete m_shader;
    delete m_program;

    m_vbo.destroy();
    m_vao.destroy();

    doneCurrent();
}

void MyGLWidget::initializeGL()
{
    m_vao.create();
    if (m_vao.isCreated())
        m_vao.bind();

    m_vbo.create();
    m_vbo.bind();
    m_vbo.allocate(...);

    m_texture = new QOpenGLTexture(QImage(...));

    m_shader = new QOpenGLShader(...);
    m_program = new QOpenGLShaderProgram(...);

    ...
}

This works for most cases, but not fully ideal as a generic solution. When the widget is reparented so that it ends up in an entirely different top-level window, something more is needed: by connecting to the aboutToBeDestroyed() signal of QOpenGLContext, cleanup can be performed whenever the OpenGL context is about to be released.

MyGLWidget::~MyGLWidget()
{
    cleanup();
}

void MyGLWidget::initializeGL()
{
    ...
    connect(context(), &QOpenGLContext::aboutToBeDestroyed, this, &MyGLWidget::cleanup);
}

void MyGLWidget::cleanup()
{
    makeCurrent();
    delete m_texture;
    m_texture = 0;
    ...
    doneCurrent();
    disconnect(context(), &QOpenGLContext::aboutToBeDestroyed, this, &MyGLWidget::cleanup);
}

Proper cleanup is especially important due to context sharing. Even though each QOpenGLWidget's associated context is destroyed together with the QOpenGLWidget, the sharable resources in that context, like textures, will stay valid until the top-level window, in which the QOpenGLWidget lived, is destroyed. Additionally, settings like Qt::AA_ShareOpenGLContexts and some Qt modules may trigger an even wider scope for sharing contexts, potentially leading to keeping the resources in question alive for the entire lifetime of the application. Therefore the safest and most robust is always to perform explicit cleanup for all resources and resource wrappers used in the QOpenGLWidget.

Limitations and Other Considerations

Putting other widgets underneath and making the QOpenGLWidget transparent will not lead to the expected results: The widgets underneath will not be visible. This is because in practice the QOpenGLWidget is drawn before all other regular, non-OpenGL widgets, and so see-through type of solutions are not feasible. Other type of layouts, like having widgets on top of the QOpenGLWidget, will function as expected.

When absolutely necessary, this limitation can be overcome by setting the Qt::WA_AlwaysStackOnTop attribute on the QOpenGLWidget. Be aware however that this breaks stacking order, for example it will not be possible to have other widgets on top of the QOpenGLWidget, so it should only be used in situations where a semi-transparent QOpenGLWidget with other widgets visible underneath is required.

Note that this does not apply when there are no other widgets underneath and the intention is to have a semi-transparent window. In that case the traditional approach of setting Qt::WA_TranslucentBackground on the top-level window is sufficient. Note that if the transparent areas are only desired in the QOpenGLWidget, then Qt::WA_NoSystemBackground will need to be turned back to false after enabling Qt::WA_TranslucentBackground. Additionally, requesting an alpha channel for the QOpenGLWidget's context via setFormat() may be necessary too, depending on the system.

QOpenGLWidget supports multiple update behaviors, just like QOpenGLWindow. In preserved mode the rendered content from the previous paintGL() call is available in the next one, allowing incremental rendering. In non-preserved mode the content is lost and paintGL() implementations are expected to redraw everything in the view.

Before Qt 5.5 the default behavior of QOpenGLWidget was to preserve the rendered contents between paintGL() calls. Since Qt 5.5 the default behavior is non-preserved because this provides better performance and the majority of applications have no need for the previous content. This also resembles the semantics of an OpenGL-based QWindow and matches the default behavior of QOpenGLWindow in that the color and ancillary buffers are invalidated for each frame. To restore the preserved behavior, call setUpdateBehavior() with PartialUpdate.

Once a QOpenGLWidget is added to a widget hierarchy, the contents of the top-level window is flushed via OpenGL-based rendering. Widgets other than the QOpenGLWidget continue to draw their content using a software-based painter, but the final composition is done through the 3D API.

Alternatives

Adding a QOpenGLWidget into a window turns on OpenGL-based compositing for the entire window. In some special cases this may not be ideal, and the old QGLWidget-style behavior with a separate, native child window is desired. Desktop applications that understand the limitations of this approach (for example when it comes to overlaps, transparency, scroll views and MDI areas), can use QOpenGLWindow with QWidget::createWindowContainer(). This is a modern alternative to QGLWidget and is faster than QOpenGLWidget due to the lack of the additional composition step. It is strongly recommended to limit the usage of this approach to cases where there is no other choice. Note that this option is not suitable for most embedded and mobile platforms, and it is known to have issues on certain desktop platforms (e.g. macOS) too. The stable, cross-platform solution is always QOpenGLWidget.

Stereoscopic rendering

Starting from 6.5 QOpenGLWidget has support for stereoscopic rendering. To enable it, set the QSurfaceFormat::StereoBuffers flag globally before the window is created, using QSurfaceFormat::SetDefaultFormat().

This will trigger paintGL() to be called twice each frame, once for each QOpenGLWidget::TargetBuffer. In paintGL(), call currentTargetBuffer() to query which one is currently being drawn to.

OpenGL is a trademark of Silicon Graphics, Inc. in the United States and other countries.