/**************************************************************************** ** ** Copyright (C) 2019 The Qt Company Ltd. ** Contact: https://www.qt.io/licensing/ ** ** This file is part of the documentation of the Qt Toolkit. ** ** $QT_BEGIN_LICENSE:FDL$ ** Commercial License Usage ** Licensees holding valid commercial Qt licenses may use this file in ** accordance with the commercial license agreement provided with the ** Software or, alternatively, in accordance with the terms contained in ** a written agreement between you and The Qt Company. For licensing terms ** and conditions see https://www.qt.io/terms-conditions. For further ** information use the contact form at https://www.qt.io/contact-us. ** ** GNU Free Documentation License Usage ** Alternatively, this file may be used under the terms of the GNU Free ** Documentation License version 1.3 as published by the Free Software ** Foundation and appearing in the file included in the packaging of ** this file. Please review the following information to ensure ** the GNU Free Documentation License version 1.3 requirements ** will be met: https://www.gnu.org/licenses/fdl-1.3.html. ** $QT_END_LICENSE$ ** ****************************************************************************/ /*! \title Qt Quick Scene Graph \page qtquick-visualcanvas-scenegraph.html \section1 The Scene Graph in Qt Quick Qt Quick 2 makes use of a dedicated scene graph that is then traversed and rendered via a graphics API such as OpenGL ES, OpenGL, Vulkan, Metal, or Direct 3D. Using a scene graph for graphics rather than the traditional imperative painting systems (QPainter and similar), means the scene to be rendered can be retained between frames and the complete set of primitives to render is known before rendering starts. This opens up for a number of optimizations, such as batch rendering to minimize state changes and discarding obscured primitives. For example, say a user-interface contains a list of ten items where each item has a background color, an icon and a text. Using the traditional drawing techniques, this would result in 30 draw calls and a similar amount of state changes. A scene graph, on the other hand, could reorganize the primitives to render such that all backgrounds are drawn in one call, then all icons, then all the text, reducing the total amount of draw calls to only 3. Batching and state change reduction like this can greatly improve performance on some hardware. The scene graph is closely tied to Qt Quick 2.0 and can not be used stand-alone. The scene graph is managed and rendered by the QQuickWindow class and custom Item types can add their graphical primitives into the scene graph through a call to QQuickItem::updatePaintNode(). The scene graph is a graphical representation of the Item scene, an independent structure that contains enough information to render all the items. Once it has been set up, it can be manipulated and rendered independently of the state of the items. On many platforms, the scene graph will even be rendered on a dedicated render thread while the GUI thread is preparing the next frame's state. \note Much of the information listed on this page is specific to the built-in, default behavior of the Qt Quick Scene graph. When using an alternative scene graph adaptation, such as, the \c software adaptation, not all concepts may apply. For more information about the different scene graph adaptations see \l{qtquick-visualcanvas-adaptations.html}{Scene Graph Adaptations}. \section1 Qt Quick Scene Graph Structure The scene graph is composed of a number of predefined node types, each serving a dedicated purpose. Although we refer to it as a scene graph, a more precise definition is node tree. The tree is built from QQuickItem types in the QML scene and internally the scene is then processed by a renderer which draws the scene. The nodes themselves do \b not contain any active drawing code nor virtual \c paint() function. Even though the node tree is mostly built internally by the existing Qt Quick QML types, it is possible for users to also add complete subtrees with their own content, including subtrees that represent 3D models. \section2 Nodes The most important node for users is the \l QSGGeometryNode. It is used to define custom graphics by defining its geometry and material. The geometry is defined using \l QSGGeometry and describes the shape or mesh of the graphical primitive. It can be a line, a rectangle, a polygon, many disconnected rectangles, or complex 3D mesh. The material defines how the pixels in this shape are filled. A node can have any number of children and geometry nodes will be rendered so they appear in child-order with parents behind their children. \note This does not say anything about the actual rendering order in the renderer. Only the visual output is guaranteed. The available nodes are: \annotatedlist{qtquick-scenegraph-nodes} Custom nodes are added to the scene graph by subclassing QQuickItem::updatePaintNode() and setting the \l {QQuickItem::ItemHasContents} flag. \warning It is crucial that native graphics (OpenGL, Vulkan, Metal, etc.) operations and interaction with the scene graph happens exclusively on the render thread, primarily during the updatePaintNode() call. The rule of thumb is to only use classes with the "QSG" prefix inside the QQuickItem::updatePaintNode() function. For more details, see the \l {Scene Graph - Custom Geometry}. \section3 Preprocessing Nodes have a virtual QSGNode::preprocess() function, which will be called before the scene graph is rendered. Node subclasses can set the flag \l QSGNode::UsePreprocess and override the QSGNode::preprocess() function to do final preparation of their node. For example, dividing a bezier curve into the correct level of detail for the current scale factor or updating a section of a texture. \section3 Node Ownership Ownership of the nodes is either done explicitly by the creator or by the scene graph by setting the flag \l QSGNode::OwnedByParent. Assigning ownership to the scene graph is often preferable as it simplifies cleanup when the scene graph lives outside the GUI thread. \section2 Materials The material describes how the interior of a geometry in a \l QSGGeometryNode is filled. It encapsulates graphics shaders for the vertex and fragment stages of the graphics pipeline and provides ample flexibility in what can be achieved, though most of the Qt Quick items themselves only use very basic materials, such as solid color and texture fills. For users who just want to apply custom shading to a QML Item type, it is possible to do this directly in QML using the \l ShaderEffect type. Below is a complete list of material classes: \annotatedlist{qtquick-scenegraph-materials} \section2 Convenience Nodes The scene graph API is low-level and focuses on performance rather than convenience. Writing custom geometries and materials from scratch, even the most basic ones, requires a non-trivial amount of code. For this reason, the API includes a few convenience classes to make the most common custom nodes readily available. \list \li \l QSGSimpleRectNode - a QSGGeometryNode subclass which defines a rectangular geometry with a solid color material. \li \l QSGSimpleTextureNode - a QSGGeometryNode subclass which defines a rectangular geometry with a texture material. \endlist \section1 Scene Graph and Rendering The rendering of the scene graph happens internally in the QQuickWindow class, and there is no public API to access it. There are, however, a few places in the rendering pipeline where the user can attach application code. This can be used to add custom scene graph content or to insert arbitrary rendering commands by directly calling the graphics API (OpenGL, Vulkan, Metal, etc.) that is in use by the scene graph. The integration points are defined by the render loop. For detailed description of how the scene graph renderer works, see \l {Qt Quick Scene Graph Default Renderer}. There are two render loop variants available: \c basic, and \c threaded. \c basic is single-threaded, while \c threaded performs scene graph rendering on a dedicated thread. Qt attempts to choose a suitable loop based on the platform and possibly the graphics drivers in use. When this is not satisfactory, or for testing purposes, the environment variable \c QSG_RENDER_LOOP can be used to force the usage of a given loop. To verify which render loop is in use, enable the \c qt.scenegraph.general \l {QLoggingCategory}{logging category}. \section2 Threaded Render Loop ('threaded') On many configurations, the scene graph rendering will happen on a dedicated render thread. This is done to increase parallelism of multi-core processors and make better use of stall times such as waiting for a blocking swap buffer call. This offers significant performance improvements, but imposes certain restrictions on where and when interaction with the scene graph can happen. The following is a simple outline of how a frame gets rendered with the threaded render loop and OpenGL. The steps are the same with other graphics APIs as well, apart from the OpenGL context specifics. \image sg-renderloop-threaded.png \list 1 \li A change occurs in the QML scene, causing \c QQuickItem::update() to be called. This can be the result of for instance an animation or user input. An event is posted to the render thread to initiate a new frame. \li The render thread prepares to draw a new frame and initiates a block on the GUI thread. \li While the render thread is preparing the new frame, the GUI thread calls QQuickItem::updatePolish() to do final touch-up of items before they are rendered. \li GUI thread is blocked. \li The QQuickWindow::beforeSynchronizing() signal is emitted. Applications can make direct connections (using Qt::DirectConnection) to this signal to do any preparation required before calls to QQuickItem::updatePaintNode(). \li Synchronization of the QML state into the scene graph. This is done by calling the QQuickItem::updatePaintNode() function on all items that have changed since the previous frame. This is the only time the QML items and the nodes in the scene graph interact. \li GUI thread block is released. \li The scene graph is rendered: \list 1 \li The QQuickWindow::beforeRendering() signal is emitted. Applications can make direct connections (using Qt::DirectConnection) to this signal to use custom graphics API calls which will then stack visually beneath the QML scene. \li Items that have specified QSGNode::UsePreprocess, will have their QSGNode::preprocess() function invoked. \li The renderer processes the nodes. \li The renderer generates states and records draw calls for the graphics API in use. \li The QQuickWindow::afterRendering() signal is emitted. Applications can make direct connections (using Qt::DirectConnection) to this signal to issue custom graphics API calls which will then stack visually over the QML scene. \li The frame is now ready. The buffers are swapped (OpenGL), or a present command is recorded and the command buffers are submitted to a graphics queue (Vulkan, Metal). QQuickWindow::frameSwapped() is emitted. \endlist \li While the render thread is rendering, the GUI is free to advance animations, process events, etc. \endlist The threaded renderer is currently used by default on Windows with Direct3D 11 and with OpenGL when using opengl32.dll, Linux excluding Mesa llvmpipe, \macos with Metal, mobile platforms, and Embedded Linux with EGLFS, and with Vulkan regardless of the platform. All this may change in future releases. It is always possible to force use of the threaded renderer by setting \c {QSG_RENDER_LOOP=threaded} in the environment. \section2 Non-threaded Render Loop ('basic') The non-threaded render loop is currently used by default on Windows with OpenGL when not using the system's standard opengl32.dll, \macos with OpenGL, and Linux with some drivers. For the latter this is mostly a precautionary measure, as not all combinations of OpenGL drivers and windowing systems have been tested. On macOS and OpenGL, the threaded render loop is not supported when building with XCode 10 (10.14 SDK) or later, since this opts in to layer-backed views on macOS 10.14. You can build with Xcode 9 (10.13 SDK) to opt out of layer-backing, in which case the threaded render loop is available and used by default. There is no such restriction with Metal. Even when using the non-threaded render loop, you should write your code as if you are using the threaded renderer, as failing to do so will make the code non-portable. The following is a simplified illustration of the frame rendering sequence in the non-threaded renderer. \image sg-renderloop-singlethreaded.png \section2 Driving Animations \section3 What does \c{Advance Animations} refer to in the above diagrams? By default, a Qt Quick animation (such, as a \l NumberAnimation) is driven by the default animation driver. This relies on basic system timers, such as QObject::startTimer(). The timer typically runs with an interval of 16 milliseconds. While this will never be fully accurate and also depends on the accuracy of timers in the underlying platform, it has the benefit of being independent of the rendering. It provides uniform results regardless of the display refresh rate and if synchronization to the display's vertical sync is active or not. This is how animations work with the \c basic render loop. In order to provide more accurate results with less stutter on-screen, independent of the render loop design (be it single threaded or multiple threads) a render loop may decide to install its own custom animation driver, and take the operation of \c advancing it into its own hands, without relying on timers. This is what the \c threaded render loop implements. In fact, it installs not one, but two animation drivers: one on the gui thread (to drive regular animations, such as \l NumberAnimation), and one on the render thread (to drive render thread animations, i.e. the \l Animator types, such as \l OpacityAnimator or \l XAnimator). Both of these are advanced during the preparation of a frame, i.e. animations are now synchronized with rendering. This makes sense due to presentation being throttled to the display's vertical sync by the underlying graphics stack. Therefore, in the diagram for the \c threaded render loop above, there is an explicit \c{Advance animations} step on both threads. For the render thread, this is trivial: as the thread is being throttled to vsync, advancing animations (for \l Animator types) in each frame as if 16.67 milliseconds had elapsed gives more accurate results than relying on a system timer. (when throttled to the vsync timing, which is \c{1000/60} milliseconds with a 60 Hz refresh rate, it is fair to assume that it has been approximately that long since the same operation was done for the previous frame) The same approach works for animations on the gui (main) thread too: due to the essential synchronization of data between the gui and render threads, the gui thread is effectively throttled to the same rate as the render thread, while still having the benefit of having less work to do, leaving more headroom for the application logic since much of the rendering preparations are now offloaded to the render thread. While the above examples used 60 frames per second, Qt Quick is prepared for other refresh rates as well: the rate is queried from the QScreen and the platform. For example, with a 144 Hz screen the interval is 6.94 ms. At the same time this is exactly what can cause trouble if vsync-based throttling is not functioning as expected, because if what the render loop thinks is happening is not matching reality, incorrect animation pacing will occur. In summary, the \c threaded render loop is expected to provide smoother animations with less stutter as long as the following conditions are met: \list \li There is exactly one window (as in QQuickWindow) on-screen. \li VSync-based throttling works as expected with the underyling graphics and display stack. \endlist \section3 What if there is no or more than one window visible? When there is no renderable window, for example because our QQuickWindow is minimized (Windows) or fully obscured (macOS), we cannot present frames, thus cannot rely on the thread "working" in lockstep with the screen refresh rate. In this case, the \c threaded render loop automatically switches over to a system timer based approach to drive animations, i.e. temporarily switching over to the mechanism the \c basic loop would use. The same is true when there are more than one QQuickWindow instances on-screen. The model presented above for advancing animations on the gui thread, enabled by its synchronization with the render thread, is not satisfactory anymore, as there are now multiple sync points with multiple render threads. (one per window.) Here falling back to the system timer based approach becomes necessary as well, because how long and often the gui thread will block is now dependent on a number of factors, including the content in the windows (are they animating? how often are they updating?) and the graphics stack behavior (how exactly does it handle two or more threads presenting with wait-for-vsync?). As we cannot guarantee being throttled to the presentation rate of the window (which window would that be, to begin with?) in a stable, cross-platform manner, advancing animations cannot be based on the rendering. This switch of animation handling mechanisms is transparent to the applications. \section3 What if vsync-based throttling is disfunctional, globally disabled, or the application disabled it itself? The \c threaded render loop relies on the graphics API implementation and/or the windowing system for throttling, for example, by requesting a swap interval of 1 in case of OpenGL (GLX, EGL, WGL), calling Present() with an interval of 1 for Direct 3D, or using the presentation mode \c FIFO with Vulkan. Some graphics drivers allow users to override this setting and turn it off, ignoring Qt's request. An example of this would be a system wide control panel of the graphics driver that allows overriding the application's settings with regards to vsync. It can also happen that a graphics stack is unable to provide proper vsync-based throttling, which can be the case in some virtual machines (mainly due to using a software rasterization based implementation of OpenGL or Vulkan). Without blocking in the swap/present operation (or some other graphics operation), such a render loop would advance animations too fast. This would be no issue with the \c basic render loop, because that always relies on system timers. With \c threaded, the behavior can vary based on the Qt version: \list \li If a system is known to be unable to provide vsync-based throttling, the only option before Qt 6.4 was to use the \c basic render loop, by manually setting \c {QSG_RENDER_LOOP=basic} in the environment before running the application. \li Starting with Qt 6.4, setting either the \c{QSG_NO_VSYNC} environment variable to a non-zero value, or the window's QSurfaceFormat::swapInterval() to \c 0 can both alleviate the problem as well: by explicitly requesting disabling vsync-based blocking, regardless of the request having any effect in practice, the \c threaded render loop can by extension recognize that relying on vsync to drive animations is futile, and it will fall back to using system timers, just as it would for more than one window. \li Even better, starting from Qt 6.4 the scenegraph also attempts to recognize using some simple heuristics that the frames are being presented "too fast", and automatically switch over to system timers if seen necessary. This means that in most cases there will be no need to do anything and applications will run animations as expected even when the default render loop is the \c threaded one. While this is transparent to applications, for troubleshooting and development purposes it is useful to know that this is logged with a \c{"Window 0x7ffc8489c3d0 is determined to have broken vsync throttling ..."} message printed when \c{QSG_INFO} or \c{qt.scenegraph.general} is enabled. This method has the downside of activating only after a small set of frames, given that it first needs to collect data to evaluate, meaning that when opening a QQuickWindow the application may still show overly fast animations for a short period of time. Additionally, it may not capture all possible vsync-broken situations. \endlist Remember however, that by design none of this helps render thread animations (the \l Animator types). In the absence of vsync-based blocking, \l{Animator}{animators} will advance incorrectly, faster than expected, even when the workarounds are activated for regular \l{Animation}{animations}. \note Be aware that the rendering loop logic and event processing on the GUI (main) thread is not necessarily unthrottled even if waiting for vsync is disabled: both render loops schedule updates for windows via QWindow::requestUpdate(). This is backed by a 5 ms GUI thread timer on most platforms, in order to give time for event processing. On some platforms, e.g. macOS, it is using platform-specific APIs (such as, CVDisplayLink) to get notified about the appropriate time to prepare a new frame, likely tied to the display's vsync in some form. This can be relevant in benchmarking and similar situations. For applications and tools attempting to perform low-level benchmarking it may be beneficial to set the \c{QT_QPA_UPDATE_IDLE_TIME} environment variable to \c 0 in order to potentially reduce idle time on the GUI thread. For normal application usage the defaults should, in most cases, be sufficient. \note When in doubt, enable the \c {qt.scenegraph.general} and \c {qt.scenegraph.time.renderloop} logging categories for troubleshooting, as these may reveal some clues as to why rendering and animations are not running at the expected pace. \section2 Custom control over rendering with QQuickRenderControl When using QQuickRenderControl, the responsibility for driving the rendering loop is transferred to the application. In this case no built-in render loop is used. Instead, it is up to the application to invoke the polish, synchronize and rendering steps at the appropriate time. It is possible to implement either a threaded or non-threaded behavior similar to the ones shown above. Additionally, applications may wish to implement and install their own QAnimationDriver in combination with QQuickRenderControl. This gives full control over driving Qt Quick animations, which can be particularly important for content that is not shown on screen, bearing no relation to the presentation rate simply because there is no presenting of the frame happening. This is optional, by default animations will advance based on the system timer. \section2 Mixing Scene Graph and the native graphics API The scene graph offers two methods for integrating application-provided graphics commands: by issuing OpenGL, Vulkan, Metal, etc. commands directly, and by creating a textured node in the scene graph. By connecting to the \l QQuickWindow::beforeRendering() and \l QQuickWindow::afterRendering() signals, applications can make OpenGL calls directly into the same context as the scene graph is rendering to. With APIs like Vulkan or Metal, applications can query native objects, such as, the scene graph's command buffer, via QSGRendererInterface, and record commands to it as they see fit. As the signal names indicate, the user can then render content either under a Qt Quick scene or over it. The benefit of integrating in this manner is that no extra framebuffer nor memory is needed to perform the rendering, and a possibly expensive texturing step is eliminated. The downside is that Qt Quick decides when to call the signals and this is the only time the OpenGL application is allowed to draw. The \l {Scene Graph - OpenGL Under QML} example gives an example on how to use these signals using OpenGL. The \l {Scene Graph - Direct3D 11 Under QML} example gives an example on how to use these signals using Direct3D. The \l {Scene Graph - Metal Under QML} example gives an example on how to use these signals using Metal. The \l {Scene Graph - Vulkan Under QML} example gives an example on how to use these signals using Vulkan. The other alternative, only available for OpenGL currently, is to create a QQuickFramebufferObject, render into it, and let it be displayed in the scene graph as a texture. The \l {Scene Graph - Rendering FBOs} example shows how this can be done. Graphics APIs other than OpenGL can also follow this approach, even though QQuickFramebufferObject does not currently support them. Creating and rendering to a texture directly with the underlying API, followed by wrapping and using this resource in a Qt Quick scene in a custom QQuickItem, is demonstrated in the \l {Scene Graph - Metal Texture Import} example. That example uses Metal, the concepts however apply to all other graphics APIs as well. \warning Starting with Qt 6.0, direct usage of the underlying graphics API must be enclosed by a call to \l QQuickWindow::beginExternalCommands() and \l QQuickWindow::endExternalCommands(). This concept may be familiar from \l QPainter::beginNativePainting(), and serves a similar purpose: it allows the Qt Quick Scene Graph to recognize that any cached state and assumptions about the state within the currently recorded render pass, if there is one, are now invalid, because the application code may have altered it by working directly with the underlying graphics API. \warning When mixing OpenGL content with scene graph rendering, it is important the application does not leave the OpenGL context in a state with buffers bound, attributes enabled, special values in the z-buffer or stencil-buffer or similar. Doing so can result in unpredictable behavior. \warning The custom rendering code must be thread aware in the sense that it should not assume being executed on the GUI (main) thread of the application. \section2 Custom Items using QPainter The QQuickItem provides a subclass, QQuickPaintedItem, which allows the users to render content using QPainter. \warning Using QQuickPaintedItem uses an indirect 2D surface to render its content, either using software rasterization or using an OpenGL framebuffer object (FBO), so the rendering is a two-step operation. First rasterize the surface, then draw the surface. Using scene graph API directly is always significantly faster. \section1 Logging Support The scene graph has support for a number of logging categories. These can be useful in tracking down both performance issues and bugs in addition to being helpful to Qt contributors. \list \li \c {qt.scenegraph.time.texture} - logs the time spent doing texture uploads \li \c {qt.scenegraph.time.compilation} - logs the time spent doing shader compilation \li \c {qt.scenegraph.time.renderer} - logs the time spent in the various steps of the renderer \li \c {qt.scenegraph.time.renderloop} - logs the time spent in the various steps of the render loop. With the \c threaded render loop this gives an insight into the time elapsed between the various frame preparation steps both on the GUI and the render thread. It can therefore also be a useful troubleshooting tool, for example, to confirm how vsync-based throttling and other low-level Qt enablers, such as QWindow::requestUpdate(), affect the rendering and presentation pipeline. \li \c {qt.scenegraph.time.glyph} - logs the time spent preparing distance field glyphs \li \c {qt.scenegraph.general} - logs general information about various parts of the scene graph and the graphics stack \li \c {qt.scenegraph.renderloop} - creates a detailed log of the various stages involved in rendering. This log mode is primarily useful for developers working on Qt. \endlist The legacy \c{QSG_INFO} environment variable is also available. Setting it to a non-zero value enables the \c{qt.scenegraph.general} category. \note When encountering graphics problems, or when in doubt which render loop or graphics API is in use, always start the application with at least \c{qt.scenegraph.general} and \c{qt.rhi.*} enabled, or \c{QSG_INFO=1} set. This will then print some essential information onto the debug output during initialization. \section1 Scene Graph Backend In addition to the public API, the scene graph has an adaptation layer which opens up the implementation to do hardware specific adaptations. This is an undocumented, internal and private plugin API, which lets hardware adaptation teams make the most of their hardware. It includes: \list \li Custom textures; specifically the implementation of QQuickWindow::createTextureFromImage and the internal representation of the texture used by \l Image and \l BorderImage types. \li Custom renderer; the adaptation layer lets the plugin decide how the scene graph is traversed and rendered, making it possible to optimize the rendering algorithm for a specific hardware or to make use of extensions which improve performance. \li Custom scene graph implementation of many of the default QML types, including its text and font rendering. \li Custom animation driver; allows the animation system to hook into the low-level display vertical refresh to get smooth rendering. \li Custom render loop; allows better control over how QML deals with multiple windows. \endlist */ /*! \title Qt Quick Scene Graph Default Renderer \page qtquick-visualcanvas-scenegraph-renderer.html This document explains how the default scene graph renderer works internally, so that one can write code that uses it in an optimal fashion, both performance and feature-wise. One does not need to understand the internals of the renderer to get good performance. However, it might help when integrating with the scene graph or to figure out why it is not possible to squeeze the maximum efficiency out of the graphics chip. \note Even in the case where every frame is unique and everything is uploaded from scratch, the default renderer will perform well. The Qt Quick items in a QML scene populate a tree of QSGNode instances. Once created, this tree is a complete description of how a certain frame should be rendered. It does not contain any references back to the Qt Quick items at all and will on most platforms be processed and rendered in a separate thread. The renderer is a self contained part of the scene graph which traverses the QSGNode tree and uses geometry defined in QSGGeometryNode and shader state defined in QSGMaterial to update the graphics state and generate draw calls. If needed, the renderer can be completely replaced using the internal scene graph back-end API. This is mostly interesting for platform vendors who wish to take advantage of non-standard hardware features. For the majority of use cases, the default renderer will be sufficient. The default renderer focuses on two primary strategies to optimize the rendering: Batching of draw calls, and retention of geometry on the GPU. \section1 Batching Whereas a traditional 2D API, such as QPainter, Cairo or Context2D, is written to handle thousands of individual draw calls per frame, OpenGL and other hardware accelerated APIs perform best when the number of draw calls is very low and state changes are kept to a minimum. \note While \c OpenGL is used as an example in the following sections, the same concepts apply to other graphics APIs as well. Consider the following use case: \image visualcanvas_list.png The simplest way of drawing this list is on a cell-by-cell basis. First, the background is drawn. This is a rectangle of a specific color. In OpenGL terms this means selecting a shader program to do solid color fills, setting up the fill color, setting the transformation matrix containing the x and y offsets and then using for instance \c glDrawArrays to draw two triangles making up the rectangle. The icon is drawn next. In OpenGL terms this means selecting a shader program to draw textures, selecting the active texture to use, setting the transformation matrix, enabling alpha-blending and then using for instance \c glDrawArrays to draw the two triangles making up the bounding rectangle of the icon. The text and separator line between cells follow a similar pattern. And this process is repeated for every cell in the list, so for a longer list, the overhead imposed by OpenGL state changes and draw calls completely outweighs the benefit that using a hardware accelerated API could provide. When each primitive is large, this overhead is negligible, but in the case of a typical UI, there are many small items which add up to a considerable overhead. The default scene graph renderer works within these limitations and will try to merge individual primitives together into batches while preserving the exact same visual result. The result is fewer OpenGL state changes and a minimal amount of draw calls, resulting in optimal performance. \section2 Opaque Primitives The renderer separates between opaque primitives and primitives which require alpha blending. By using OpenGL's Z-buffer and giving each primitive a unique z position, the renderer can freely reorder opaque primitives without any regard for their location on screen and which other elements they overlap with. By looking at each primitive's material state, the renderer will create opaque batches. From Qt Quick core item set, this includes Rectangle items with opaque colors and fully opaque images, such as JPEGs or BMPs. Another benefit of using opaque primitives is that opaque primitives do not require \c GL_BLEND to be enabled, which can be quite costly, especially on mobile and embedded GPUs. Opaque primitives are rendered in a front-to-back manner with \c glDepthMask and \c GL_DEPTH_TEST enabled. On GPUs that internally do early-z checks, this means that the fragment shader does not need to run for pixels or blocks of pixels that are obscured. Beware that the renderer still needs to take these nodes into account and the vertex shader is still run for every vertex in these primitives, so if the application knows that something is fully obscured, the best thing to do is to explicitly hide it using Item::visible or Item::opacity. \note The Item::z is used to control an Item's stacking order relative to its siblings. It has no direct relation to the renderer and OpenGL's Z-buffer. \section2 Alpha Blended Primitives Once opaque primitives have been drawn, the renderer will disable \c glDepthMask, enable \c GL_BLEND and render all alpha blended primitives in a back-to-front manner. Batching of alpha blended primitives requires a bit more effort in the renderer as elements that are overlapping need to be rendered in the correct order for alpha blending to look correct. Relying on the Z-buffer alone is not enough. The renderer does a pass over all alpha blended primitives and will look at their bounding rect in addition to their material state to figure out which elements can be batched and which can not. \image visualcanvas_overlap.png In the left-most case, the blue backgrounds can be drawn in one call and the two text elements in another call, as the texts only overlap a background which they are stacked in front of. In the right-most case, the background of "Item 4" overlaps the text of "Item 3" so in this case, each of backgrounds and texts needs to be drawn using separate calls. Z-wise, the alpha primitives are interleaved with the opaque nodes and may trigger early-z when available, but again, setting Item::visible to false is always faster. \section2 Mixing with 3D Primitives The scene graph can support pseudo 3D and proper 3D primitives. For instance, one can implement a "page curl" effect using a ShaderEffect or implement a bumpmapped torus using QSGGeometry and a custom material. While doing so, one needs to take into account that the default renderer already makes use of the depth buffer. The renderer modifies the vertex shader returned from QSGMaterialShader::vertexShader() and compresses the z values of the vertex after the model-view and projection matrices have been applied and then adds a small translation on the z to position it the correct z position. The compression assumes that the z values are in the range of 0 to 1. \section2 Texture Atlas The active texture is a unique OpenGL state, which means that multiple primitives using different OpenGL textures cannot be batched. The Qt Quick scene graph, for this reason, allows multiple QSGTexture instances to be allocated as smaller sub-regions of a larger texture; a texture atlas. The biggest benefit of texture atlases is that multiple QSGTexture instances now refer to the same OpenGL texture instance. This makes it possible to batch textured draw calls as well, such as Image items, BorderImage items, ShaderEffect items and also C++ types such as QSGSimpleTextureNode and custom QSGGeometryNodes using textures. \note Large textures do not go into the texture atlas. Atlas based textures are created by passing QQuickWindow::TextureCanUseAtlas to the QQuickWindow::createTextureFromImage(). \note Atlas based textures do not have texture coordinates ranging from 0 to 1. Use QSGTexture::normalizedTextureSubRect() to get the atlas texture coordinates. The scene graph uses heuristics to figure out how large the atlas should be and what the size threshold for being entered into the atlas is. If different values are needed, it is possible to override them using the environment variables \c {QSG_ATLAS_WIDTH=[width]}, \c {QSG_ATLAS_HEIGHT=[height]} and \c {QSG_ATLAS_SIZE_LIMIT=[size]}. Changing these values will mostly be interesting for platform vendors. \section1 Batch Roots In addition to merging compatible primitives into batches, the default renderer also tries to minimize the amount of data that needs to be sent to the GPU for every frame. The default renderer identifies subtrees which belong together and tries to put these into separate batches. Once batches are identified, they are merged, uploaded and stored in GPU memory, using Vertex Buffer Objects. \section2 Transform Nodes Each Qt Quick Item inserts a QSGTransformNode into the scene graph tree to manage its x, y, scale or rotation. Child items will be populated under this transform node. The default renderer tracks the state of transform nodes between frames and will look at subtrees to decide if a transform node is a good candidate to become a root for a set of batches. A transform node which changes between frames and which has a fairly complex subtree can become a batch root. QSGGeometryNodes in the subtree of a batch root are pre-transformed relative to the root on the CPU. They are then uploaded and retained on the GPU. When the transform changes, the renderer only needs to update the matrix of the root, not each individual item, making list and grid scrolling very fast. For successive frames, as long as nodes are not being added or removed, rendering the list is effectively for free. When new content enters the subtree, the batch that gets it is rebuilt, but this is still relatively fast. There are usually several unchanging frames for every frame with added or removed nodes when panning through a grid or list. Another benefit of identifying transform nodes as batch roots is that it allows the renderer to retain the parts of the tree that have not changed. For instance, say a UI consists of a list and a button row. When the list is being scrolled and delegates are being added and removed, the rest of the UI, the button row, is unchanged and can be drawn using the geometry already stored on the GPU. The node and vertex threshold for a transform node to become a batch root can be overridden using the environment variables \c {QSG_RENDERER_BATCH_NODE_THRESHOLD=[count]} and \c {QSG_RENDERER_BATCH_VERTEX_THRESHOLD=[count]}. Overriding these flags will be mostly useful for platform vendors. \note Beneath a batch root, one batch is created for each unique set of material state and geometry type. \section2 Clipping When setting Item::clip to true, it will create a QSGClipNode with a rectangle in its geometry. The default renderer will apply this clip by using scissoring in OpenGL. If the item is rotated by a non-90-degree angle, the OpenGL's stencil buffer is used. Qt Quick Item only supports setting a rectangle as clip through QML, but the scene graph API and the default renderer can use any shape for clipping. When applying a clip to a subtree, that subtree needs to be rendered with a unique OpenGL state. This means that when Item::clip is true, batching of that item is limited to its children. When there are many children, like a ListView or GridView, or complex children, like a TextArea, this is fine. One should, however, use clip on smaller items with caution as it prevents batching. This includes button label, text field or list delegate and table cells. Clipping a Flickable (or item view) can often be avoided by arranging the UI so that opaque items cover areas around the Flickable, and otherwise relying on the window edges to clip everything else. Setting Item::clip to \c true also sets the \l QQuickItem::ItemIsViewport flag; child items with the \l QQuickItem::ItemObservesViewport flag may use the viewport for a rough pre-clipping step: e.g. \l Text omits lines of text that are completely outside the viewport. Omitting scene graph nodes or limiting the \l {QSGGeometry::vertexCount()}{vertices} is an optimization, which can be achieved by setting the \l {QQuickItem::flags()}{flags} in C++ rather than setting \l Item::clip in QML. When implementing QQuickItem::updatePaintNode() in a custom item, if it can render a lot of details over a large geometric area, you should think about whether it's efficient to limit the graphics to the viewport; if so, you can set the \l {QQuickItem::} {ItemObservesViewport} flag and read the currently exposed area from QQuickItem::clipRect(). One consequence is that updatePaintNode() will be called more often (typically once per frame whenever content is moving in the viewport). \section2 Vertex Buffers Each batch uses a vertex buffer object (VBO) to store its data on the GPU. This vertex buffer is retained between frames and updated when the part of the scene graph that it represents changes. By default, the renderer will upload data into the VBO using \c GL_STATIC_DRAW. It is possible to select different upload strategy by setting the environment variable \c {QSG_RENDERER_BUFFER_STRATEGY=[strategy]}. Valid values are \c stream and \c dynamic. Changing this value is mostly useful for platform vendors. \section1 Antialiasing The scene graph supports two types of antialiasing. By default, primitives such as rectangles and images will be antialiased by adding more vertices along the edge of the primitives so that the edges fade to transparent. We call this method \e {vertex antialiasing}. If the user requests a multisampled OpenGL context, by setting a QSurfaceFormat with samples greater than \c 0 using QQuickWindow::setFormat(), the scene graph will prefer multisample based antialiasing (MSAA). The two techniques will affect how the rendering happens internally and have different limitations. It is also possible to override the antialiasing method used by setting the environment variable \c {QSG_ANTIALIASING_METHOD} to either \c vertex or \c {msaa}. Vertex antialiasing can produce seams between edges of adjacent primitives, even when the two edges are mathematically the same. Multisample antialiasing does not. \section2 Vertex Antialiasing Vertex antialiasing can be enabled and disabled on a per-item basis using the Item::antialiasing property. It will work regardless of what the underlying hardware supports and produces higher quality antialiasing, both for normally rendered primitives and also for primitives captured into framebuffer objects, for instance using the ShaderEffectSource type. The downside to using vertex antialiasing is that each primitive with antialiasing enabled will have to be blended. In terms of batching, this means that the renderer needs to do more work to figure out if the primitive can be batched or not and due to overlaps with other elements in the scene, it may also result in less batching, which could impact performance. On low-end hardware blending can also be quite expensive so for an image or rounded rectangle that covers most of the screen, the amount of blending needed for the interior of these primitives can result in significant performance loss as the entire primitive must be blended. \section2 Multisample Antialiasing Multisample antialiasing is a hardware feature where the hardware calculates a coverage value per pixel in the primitive. Some hardware can multisample at a very low cost, while other hardware may need both more memory and more GPU cycles to render a frame. Using multisample antialiasing, many primitives, such as rounded rectangles and image elements can be antialiased and still be \e opaque in the scene graph. This means the renderer has an easier job when creating batches and can rely on early-z to avoid overdraw. When multisample antialiasing is used, content rendered into framebuffer objects need additional extensions to support multisampling of framebuffers. Typically \c GL_EXT_framebuffer_multisample and \c GL_EXT_framebuffer_blit. Most desktop chips have these extensions present, but they are less common in embedded chips. When framebuffer multisampling is not available in the hardware, content rendered into framebuffer objects will not be antialiased, including the content of a ShaderEffectSource. \section1 Performance As stated in the beginning, understanding the finer details of the renderer is not required to get good performance. It is written to optimize for common use cases and will perform quite well under almost any circumstances. \list \li Good performance comes from effective batching, with as little as possible of the geometry being uploaded again and again. By setting the environment variable \c {QSG_RENDERER_DEBUG=render}, the renderer will output statistics on how well the batching goes, how many batches are used, which batches are retained and which are opaque and not. When striving for optimal performance, uploads should happen only when really needed, batches should be fewer than 10 and at least 3-4 of them should be opaque. \li The default renderer does not do any CPU-side viewport clipping nor occlusion detection. If something is not supposed to be visible, it should not be shown. Use \c {Item::visible: false} for items that should not be drawn. The primary reason for not adding such logic is that it adds additional cost which would also hurt applications that took care in behaving well. \li Make sure the texture atlas is used. The Image and BorderImage items will use it unless the image is too large. For textures created in C++, pass QQuickWindow::TextureCanUseAtlas when calling QQuickWindow::createTexture(). By setting the environment variable \c {QSG_ATLAS_OVERLAY} all atlas textures will be colorized so they are easily identifiable in the application. \li Use opaque primitives where possible. Opaque primitives are faster to process in the renderer and faster to draw on the GPU. For instance, PNG files will often have an alpha channel, even though each pixel is fully opaque. JPG files are always opaque. When providing images to a QQuickImageProvider or creating images with QQuickWindow::createTextureFromImage(), let the image have QImage::Format_RGB32, when possible. \li Be aware of that overlapping compound items, like in the illustration above, cannot be batched. \li Clipping breaks batching. Never use on a per-item basis, inside table cells, item delegates or similar. Instead of clipping text, use eliding. Instead of clipping an image, create a QQuickImageProvider that returns a cropped image. \li Batching only works for 16-bit indices. All built-in items use 16-bit indices, but a custom geometry is free to also use 32-bit indices. \li Some material flags prevent batching, the most limiting one being QSGMaterial::RequiresFullMatrix which prevents all batching. \li Applications with a monochrome background should set it using QQuickWindow::setColor() rather than using a top-level Rectangle item. QQuickWindow::setColor() will be used in a call to \c glClear(), which is potentially faster. \li Mipmapped Image items are not placed in the global atlas and will not be batched. \li A bug in the OpenGL driver related to framebuffer object (FBO) readbacks may corrupt rendered glyphs. If you set the \c QML_USE_GLYPHCACHE_WORKAROUND environment variable, Qt keeps an additional copy of the glyph in RAM. This means that performance is slightly lower when drawing glyphs that have not been drawn before, as Qt accesses the extra copy via the CPU. It also means that the glyph cache will use twice as much memory. The quality is not affected by this. \endlist If an application performs poorly, make sure that rendering is actually the bottleneck. Use a profiler! The environment variable \c {QSG_RENDER_TIMING=1} will output a number of useful timing parameters which can be useful in pinpointing where a problem lies. \section1 Visualizing To visualize the various aspects of the scene graph's default renderer, the \c QSG_VISUALIZE environment variable can be set to one of the values detailed in each section below. We provide examples of the output of some of the variables using the following QML code: \code import QtQuick 2.2 Rectangle { width: 200 height: 140 ListView { id: clippedList x: 20 y: 20 width: 70 height: 100 clip: true model: ["Item A", "Item B", "Item C", "Item D"] delegate: Rectangle { color: "lightblue" width: parent.width height: 25 Text { text: modelData anchors.fill: parent horizontalAlignment: Text.AlignHCenter verticalAlignment: Text.AlignVCenter } } } ListView { id: clippedDelegateList x: clippedList.x + clippedList.width + 20 y: 20 width: 70 height: 100 clip: true model: ["Item A", "Item B", "Item C", "Item D"] delegate: Rectangle { color: "lightblue" width: parent.width height: 25 clip: true Text { text: modelData anchors.fill: parent horizontalAlignment: Text.AlignHCenter verticalAlignment: Text.AlignVCenter } } } } \endcode For the ListView on the left, we set its \l {Item::clip}{clip} property to \c true. For the ListView on right, we also set each delegate's \l {Item::clip}{clip} property to \c true to illustrate the effects of clipping on batching. \image visualize-original.png "Original" Original \note The visualized elements do not respect clipping, and rendering order is arbitrary. \section2 Visualizing Batches Setting \c QSG_VISUALIZE to \c batches visualizes batches in the renderer. Merged batches are drawn with a solid color and unmerged batches are drawn with a diagonal line pattern. Few unique colors means good batching. Unmerged batches are bad if they contain many individual nodes. \image visualize-batches.png "batches" \c QSG_VISUALIZE=batches \section2 Visualizing Clipping Setting \c QSG_VISUALIZE to \c clip draws red areas on top of the scene to indicate clipping. As Qt Quick Items do not clip by default, no clipping is usually visualized. \image visualize-clip.png \c QSG_VISUALIZE=clip \section2 Visualizing Changes Setting \c QSG_VISUALIZE to \c changes visualizes changes in the renderer. Changes in the scenegraph are visualized with a flashing overlay of a random color. Changes on a primitive are visualized with a solid color, while changes in an ancestor, such as matrix or opacity changes, are visualized with a pattern. \section2 Visualizing Overdraw Setting \c QSG_VISUALIZE to \c overdraw visualizes overdraw in the renderer. Visualize all items in 3D to highlight overdraws. This mode can also be used to detect geometry outside the viewport to some extent. Opaque items are rendered with a green tint, while translucent items are rendered with a red tint. The bounding box for the viewport is rendered in blue. Opaque content is easier for the scenegraph to process and is usually faster to render. Note that the root rectangle in the code above is superfluous as the window is also white, so drawing the rectangle is a waste of resources in this case. Changing it to an Item can give a slight performance boost. \image visualize-overdraw-1.png "overdraw-1" \image visualize-overdraw-2.png "overdraw-2" \c QSG_VISUALIZE=overdraw \section1 Rendering via the Qt Rendering Hardware Interface From Qt 6.0 onwards, the default adaptation always renders via a graphics abstraction layer, the Qt Rendering Hardware Interface (RHI), provided by the \l QtGui module. This means that, unlike Qt 5, no direct OpenGL calls are made by the scene graph. Rather, it records resource and draw commands by using the RHI APIs, which then translate the command stream into OpenGL, Vulkan, Metal, or Direct 3D calls. Shader handling is also unified by writing shader code once, compiling to \l{https://www.khronos.org/spir/}{SPIR-V}, and then translating to the language appropriate for the various graphics APIs. To control the behavior, the following environment variables can be used: \table 100% \header \li Environment Variable \li Possible Values \li Description \row \li \c QSG_RHI_BACKEND \li \c vulkan, \c metal, \c opengl, \c d3d11 \li Requests the specific RHI backend. By default the targeted graphics API is chosen based on the platform, unless overridden by this variable or the equivalent C++ APIs. The defaults are currently Direct3D 11 for Windows, Metal for macOS, OpenGL elsewhere. \row \li \c QSG_INFO \li \c 1 \li Like with the OpenGL-based rendering path, setting this enables printing system information when initializing the Qt Quick scene graph. This can be very useful for troubleshooting. \row \li \c QSG_RHI_DEBUG_LAYER \li \c 1 \li Where applicable (Vulkan, Direct3D), enables the graphics API implementation's debug and/or validation layers, if available. \row \li \c QSG_RHI_PREFER_SOFTWARE_RENDERER \li \c 1 \li Requests choosing an adapter or physical device that uses software-based rasterization. Applicable only when the underlying API has support for enumerating adapters (for example, Direct3D or Vulkan), and is ignored otherwise. \endtable Applications wishing to always run with a single given graphics API, can request this via C++ as well. For example, the following call made early in main(), before constructing any QQuickWindow, forces the use of Vulkan (and will fail otherwise): \badcode QQuickWindow::setGraphicsApi(QSGRendererInterface::Vulkan); \endcode See QSGRendererInterface::GraphicsApi. The enum values \c OpenGL, \c Vulkan, \c Metal, \c Direct3D11 are equivalent in effect to running with \c QSG_RHI_BACKEND set to the equivalent string key. All QRhi backends will choose the system default GPU adapter or physical device, unless overridden by \c{QSG_RHI_PREFER_SOFTWARE_RENDERER} or a backend-specific variable, such as, \c{QT_D3D_ADAPTER_INDEX} or \c{QT_VK_PHYSICAL_DEVICE_INDEX}. No further adapter configurability is provided at this time. */