The QVulkanWindow class is a convenience subclass of QWindow to perform Vulkan rendering. More...
Header: | #include <QVulkanWindow> |
qmake: | QT += gui |
Since: | Qt 5.10 |
Inherits: | QWindow |
QVulkanWindow(QWindow *parent = nullptr) | |
virtual | ~QVulkanWindow() |
int | availablePhysicalDevices() |
QMatrix4x4 | clipCorrectionMatrix() |
int | colorFormat() const |
int | concurrentFrameCount() const |
virtual QVulkanWindowRenderer * | createRenderer() |
int | currentCommandBuffer() const |
int | currentFrame() const |
int | currentFramebuffer() const |
int | currentSwapChainImageIndex() const |
int | defaultRenderPass() const |
int | depthStencilFormat() const |
VkImage | depthStencilImage() const |
VkImageView | depthStencilImageView() const |
int | device() const |
int | deviceLocalMemoryIndex() const |
QVulkanWindow::Flags | flags() const |
void | frameReady() |
QImage | grab() |
int | graphicsCommandPool() const |
int | graphicsQueue() const |
int | hostVisibleMemoryIndex() const |
bool | isValid() const |
VkImage | msaaColorImage(int idx) const |
VkImageView | msaaColorImageView(int idx) const |
int | physicalDevice() const |
const int * | physicalDeviceProperties() const |
int | sampleCountFlagBits() const |
void | setDeviceExtensions(const QByteArrayList &extensions) |
void | setFlags(QVulkanWindow::Flags flags) |
void | setPhysicalDeviceIndex(int idx) |
void | setPreferredColorFormats(const int &formats) |
void | setSampleCount(int sampleCount) |
QVulkanInfoVector<QVulkanExtension> | supportedDeviceExtensions() |
QVector<int> | supportedSampleCounts() |
bool | supportsGrab() const |
VkImage | swapChainImage(int idx) const |
int | swapChainImageCount() const |
QSize | swapChainImageSize() const |
VkImageView | swapChainImageView(int idx) const |
void | frameGrabbed(const QImage &image) |
const int | MAX_CONCURRENT_FRAME_COUNT |
The QVulkanWindow class is a convenience subclass of QWindow to perform Vulkan rendering.
QVulkanWindow is a Vulkan-capable QWindow that manages a Vulkan device, a graphics queue, a command pool and buffer, a depth-stencil image and a double-buffered FIFO swapchain, while taking care of correct behavior when it comes to events like resize, special situations like not having a device queue supporting both graphics and presentation, device lost scenarios, and additional functionality like reading the rendered content back. Conceptually it is the counterpart of QOpenGLWindow in the Vulkan world.
Note: QVulkanWindow does not always eliminate the need to implement a fully custom QWindow subclass as it will not necessarily be sufficient in advanced use cases.
QVulkanWindow can be embedded into QWidget-based user interfaces via QWidget::createWindowContainer(). This approach has a number of limitations, however. Make sure to study the documentation first.
A typical application using QVulkanWindow may look like the following:
class VulkanRenderer : public QVulkanWindowRenderer { public: VulkanRenderer(QVulkanWindow *w) : m_window(w) { } void initResources() override { m_devFuncs = m_window->vulkanInstance()->deviceFunctions(m_window->device()); ... } void initSwapChainResources() override { ... } void releaseSwapChainResources() override { ... } void releaseResources() override { ... } void startNextFrame() override { VkCommandBuffer cmdBuf = m_window->currentCommandBuffer(); ... m_devFuncs->vkCmdBeginRenderPass(...); ... m_window->frameReady(); } private: QVulkanWindow *m_window; QVulkanDeviceFunctions *m_devFuncs; }; class VulkanWindow : public QVulkanWindow { public: QVulkanWindowRenderer *createRenderer() override { return new VulkanRenderer(this); } }; int main(int argc, char *argv[]) { QGuiApplication app(argc, argv); QVulkanInstance inst; // enable the standard validation layers, when available inst.setLayers(QByteArrayList() << "VK_LAYER_LUNARG_standard_validation"); if (!inst.create()) qFatal("Failed to create Vulkan instance: %d", inst.errorCode()); VulkanWindow w; w.setVulkanInstance(&inst); w.showMaximized(); return app.exec(); }
As it can be seen in the example, the main patterns in QVulkanWindow usage are:
For troubleshooting, enable the logging category qt.vulkan
. Critical errors are printed via qWarning() automatically.
There are two notable differences to be aware of: First, with Vulkan Y points down the screen in clip space, while OpenGL uses an upwards pointing Y axis. Second, the standard OpenGL projection matrix assume a near and far plane values of -1 and 1, while Vulkan prefers 0 and 1.
In order to help applications migrate from OpenGL-based code without having to flip Y coordinates in the vertex data, and to allow using QMatrix4x4 functions like QMatrix4x4::perspective() while keeping the Vulkan viewport's minDepth and maxDepth set to 0 and 1, QVulkanWindow provides a correction matrix retrievable by calling clipCorrectionMatrix().
While disabled by default, multisample antialiasing is fully supported by QVulkanWindow. Additional color buffers and resolving into the swapchain's non-multisample buffers are all managed automatically.
To query the supported sample counts, call supportedSampleCounts(). When the returned set contains 4, 8, ..., passing one of those values to setSampleCount() requests multisample rendering.
Note: unlike QSurfaceFormat::setSamples(), the list of supported sample counts are exposed to the applications in advance and there is no automatic falling back to lower sample counts in setSampleCount(). If the requested value is not supported, a warning is shown and a no multisampling will be used.
When supportsGrab() returns true, QVulkanWindow can perform readbacks from the color buffer into a QImage. grab() is a slow and inefficient operation, so frequent usage should be avoided. It is nonetheless valuable since it allows applications to take screenshots, or tools and tests to process and verify the output of the GPU rendering.
While many applications will be fine with the default behavior of QVulkanWindow when it comes to swapchain image formats, setPreferredColorFormats() allows requesting a pre-defined format. This is useful most notably when working in the sRGB color space. Passing a format like
VK_FORMAT_B8G8R8A8_SRGB
results in choosing an sRGB format, when available.
During application development it can be extremely valuable to have the Vulkan validation layers enabled. As shown in the example code above, calling QVulkanInstance::setLayers() on the QVulkanInstance before QVulkanInstance::create() enables validation, assuming the Vulkan driver stack in the system contains the necessary layers.
Note: Be aware of platform-specific differences. On desktop platforms installing the Vulkan SDK is typically sufficient. However, Android for example requires deploying additional shared libraries together with the application, and also mandates a different list of validation layer names. See the Android Vulkan development pages for more information.
Note: QVulkanWindow does not expose device layers since this functionality has been deprecated since version 1.0.13 of the Vulkan API.
See also QVulkanInstance and QWindow.
This enum describes the flags that can be passed to setFlags().
Constant | Value | Description |
---|---|---|
QVulkanWindow::PersistentResources |
0x01 |
Ensures no graphics resources are released when the window becomes unexposed. The default behavior is to release everything, and reinitialize later when becoming visible again. |
The Flags type is a typedef for QFlags<Flag>. It stores an OR combination of Flag values.
Constructs a new QVulkanWindow with the given parent.
The surface type is set to QSurface::VulkanSurface.
[virtual]
QVulkanWindow::~QVulkanWindow()Destructor.
Returns the list of properties for the supported physical devices in the system.
Note: This function can be called before making the window visible.
Returns a QMatrix4x4 that can be used to correct for coordinate system differences between OpenGL and Vulkan.
By pre-multiplying the projection matrix with this matrix, applications can continue to assume OpenGL-style Y coordinates in clip space (i.e. Y pointing upwards), and can set minDepth and maxDepth to 0 and 1, respectively, without any further corrections to the vertex Z positions, while using the projection matrices retrieved from the QMatrix4x4 functions, such as QMatrix4x4::perspective(), as-is.
Note: With vertex data following the default OpenGL rules (that is, the front face being CCW), the correct winding order in the rasterization state after applying this matrix is clockwise
(VK_FRONT_FACE_CLOCKWISE
).
Returns the color buffer format used by the swapchain.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initResources() up until QVulkanWindowRenderer::releaseResources().
See also setPreferredColorFormats().
Returns the number of frames that can be potentially active at the same time.
Note: The value is constant for the entire lifetime of the QVulkanWindow.
class Renderer { ... VkDescriptorBufferInfo m_uniformBufInfo[QVulkanWindow::MAX_CONCURRENT_FRAME_COUNT]; }; void Renderer::startNextFrame() { const int count = m_window->concurrentFrameCount(); for (int i = 0; i < count; ++i) m_uniformBufInfo[i] = ... ... }
See also currentFrame().
[virtual]
QVulkanWindowRenderer
*QVulkanWindow::createRenderer()Returns a new instance of QVulkanWindowRenderer.
This virtual function is called once during the lifetime of the window, at some point after making it visible for the first time.
The default implementation returns null and so no rendering will be performed apart from clearing the buffers.
The window takes ownership of the returned renderer object.
Returns The active command buffer for the current swap chain image. Implementations of QVulkanWindowRenderer::startNextFrame() are expected to add commands to this command buffer.
Note: This function must only be called from within startNextFrame() and, in case of asynchronous command generation, up until the call to frameReady().
Returns the current frame index in the range [0, concurrentFrameCount() - 1].
Renderer implementations will have to ensure that uniform data and other dynamic resources exist in multiple copies, in order to prevent frame N altering the data used by the still-active frames N - 1, N - 2, ... N - concurrentFrameCount() + 1.
To avoid relying on dynamic array sizes, applications can use MAX_CONCURRENT_FRAME_COUNT when declaring arrays. This is guaranteed to be always equal to or greater than the value returned from concurrentFrameCount(). Such arrays can then be indexed by the value returned from this function.
class Renderer { ... VkDescriptorBufferInfo m_uniformBufInfo[QVulkanWindow::MAX_CONCURRENT_FRAME_COUNT]; }; void Renderer::startNextFrame() { VkDescriptorBufferInfo &uniformBufInfo(m_uniformBufInfo[m_window->currentFrame()]); ... }
Note: This function must only be called from within startNextFrame() and, in case of asynchronous command generation, up until the call to frameReady().
See also concurrentFrameCount().
Returns a VkFramebuffer for the current swapchain image using the default render pass.
The framebuffer has two attachments (color, depth-stencil) when multisampling is not in use, and three (color resolve, depth-stencil, multisample color) when sampleCountFlagBits() is greater than VK_SAMPLE_COUNT_1_BIT
. Renderers must take this into account, for example when providing clear values.
Note: Applications are not required to use this framebuffer in case they provide their own render pass instead of using the one returned from defaultRenderPass().
Note: This function must only be called from within startNextFrame() and, in case of asynchronous command generation, up until the call to frameReady().
See also defaultRenderPass().
Returns the current swap chain image index in the range [0, swapChainImageCount() - 1].
Note: This function must only be called from within startNextFrame() and, in case of asynchronous command generation, up until the call to frameReady().
Returns a typical render pass with one sub-pass.
Note: Applications are not required to use this render pass. However, they are then responsible for ensuring the current swap chain and depth-stencil images get transitioned from VK_IMAGE_LAYOUT_UNDEFINED
to VK_IMAGE_LAYOUT_PRESENT_SRC_KHR
and VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL
either via the application's custom render pass or by other means.
Note: Stencil read/write is not enabled in this render pass.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initResources() up until QVulkanWindowRenderer::releaseResources().
See also currentFramebuffer().
Returns the format used by the depth-stencil buffer(s).
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initResources() up until QVulkanWindowRenderer::releaseResources().
Returns the depth-stencil image.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initSwapChainResources() up until QVulkanWindowRenderer::releaseSwapChainResources().
Returns the depth-stencil image view.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initSwapChainResources() up until QVulkanWindowRenderer::releaseSwapChainResources().
Returns the active logical device.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initResources() up until QVulkanWindowRenderer::releaseResources().
Returns a device local memory type index suitable for general use.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initResources() up until QVulkanWindowRenderer::releaseResources().
Return the requested flags.
See also setFlags().
[signal]
void QVulkanWindow::frameGrabbed(const QImage &image)This signal is emitted when the image is ready.
This function must be called exactly once in response to each invocation of the QVulkanWindowRenderer::startNextFrame() implementation. At the time of this call, the main command buffer, exposed via currentCommandBuffer(), must have all necessary rendering commands added to it since this function will trigger submitting the commands and queuing the present command.
Note: This function must only be called from the gui/main thread, which is where QVulkanWindowRenderer's functions are invoked and where the QVulkanWindow instance lives.
See also QVulkanWindowRenderer::startNextFrame().
Builds and renders the next frame without presenting it, then performs a blocking readback of the image content.
Returns the image if the renderer's startNextFrame() implementation calls back frameReady() directly. Otherwise, returns an incomplete image, that has the correct size but not the content yet. The content will be delivered via the frameGrabbed() signal in the latter case.
Note: This function should not be called when a frame is in progress (that is, frameReady() has not yet been called back by the application).
Note: This function is potentially expensive due to the additional, blocking readback.
Note: This function currently requires that the swapchain supports usage as a transfer source (VK_IMAGE_USAGE_TRANSFER_SRC_BIT
), and will fail otherwise.
Returns the active graphics command pool.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initResources() up until QVulkanWindowRenderer::releaseResources().
Returns the active graphics queue.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initResources() up until QVulkanWindowRenderer::releaseResources().
Returns a host visible memory type index suitable for general use.
The returned memory type will be both host visible and coherent. In addition, it will also be cached, if possible.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initResources() up until QVulkanWindowRenderer::releaseResources().
Returns true if this window has successfully initialized all Vulkan resources, including the swapchain.
Note: Initialization happens on the first expose event after the window is made visible.
Returns the specified multisample color image, or VK_NULL_HANDLE
if multisampling is not in use.
idx must be in the range [0, swapChainImageCount() - 1].
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initSwapChainResources() up until QVulkanWindowRenderer::releaseSwapChainResources().
Returns the specified multisample color image view, or VK_NULL_HANDLE
if multisampling is not in use.
idx must be in the range [0, swapChainImageCount() - 1].
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initSwapChainResources() up until QVulkanWindowRenderer::releaseSwapChainResources().
Returns the active physical device.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::preInitResources() up until QVulkanWindowRenderer::releaseResources().
Returns a pointer to the properties for the active physical device.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::preInitResources() up until QVulkanWindowRenderer::releaseResources().
Returns the current sample count as a VkSampleCountFlagBits
value.
When targeting the default render target, the rasterizationSamples
field of VkPipelineMultisampleStateCreateInfo
must be set to this value.
See also setSampleCount() and supportedSampleCounts().
Sets the list of device extensions to be enabled.
Unsupported extensions are ignored.
The swapchain extension will always be added automatically, no need to include it in this list.
Note: This function must be called before the window is made visible or at latest in QVulkanWindowRenderer::preInitResources(), and has no effect if called afterwards.
Configures the behavior based on the provided flags.
Note: This function must be called before the window is made visible or at latest in QVulkanWindowRenderer::preInitResources(), and has no effect if called afterwards.
See also flags().
Requests the usage of the physical device with index idx. The index corresponds to the list returned from availablePhysicalDevices().
By default the first physical device is used.
Note: This function must be called before the window is made visible or at latest in QVulkanWindowRenderer::preInitResources(), and has no effect if called afterwards.
Sets the preferred formats of the swapchain.
By default no application-preferred format is set. In this case the surface's preferred format will be used or, in absence of that, VK_FORMAT_B8G8R8A8_UNORM
.
The list in formats is ordered. If the first format is not supported, the second will be considered, and so on. When no formats in the list are supported, the behavior is the same as in the default case.
To query the actual format after initialization, call colorFormat().
Note: This function must be called before the window is made visible or at latest in QVulkanWindowRenderer::preInitResources(), and has no effect if called afterwards.
Note: Reimplementing QVulkanWindowRenderer::preInitResources() allows dynamically examining the list of supported formats, should that be desired. There the surface is retrievable via QVulkanInstace::surfaceForWindow(), while this function can still safely be called to affect the later stages of initialization.
See also colorFormat().
Requests multisample antialiasing with the given sampleCount. The valid values are 1, 2, 4, 8, ... up until the maximum value supported by the physical device.
When the sample count is greater than 1, QVulkanWindow will create a multisample color buffer instead of simply targeting the swapchain's images. The rendering in the multisample buffer will get resolved into the non-multisample buffers at the end of each frame.
To examine the list of supported sample counts, call supportedSampleCounts().
When setting up the rendering pipeline, call sampleCountFlagBits() to query the active sample count as a VkSampleCountFlagBits
value.
Note: This function must be called before the window is made visible or at latest in QVulkanWindowRenderer::preInitResources(), and has no effect if called afterwards.
See also supportedSampleCounts() and sampleCountFlagBits().
Returns the list of the extensions that are supported by logical devices created from the physical device selected by setPhysicalDeviceIndex().
Note: This function can be called before making the window visible.
Returns the set of supported sample counts when using the physical device selected by setPhysicalDeviceIndex(), as a sorted vector.
By default QVulkanWindow uses a sample count of 1. By calling setSampleCount() with a different value (2, 4, 8, ...) from the set returned by this function, multisample anti-aliasing can be requested.
Note: This function can be called before making the window visible.
See also setSampleCount().
Returns true if the swapchain supports usage as transfer source, meaning grab() is functional.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initSwapChainResources() up until QVulkanWindowRenderer::releaseSwapChainResources().
Returns the specified swap chain image.
idx must be in the range [0, swapChainImageCount() - 1].
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initSwapChainResources() up until QVulkanWindowRenderer::releaseSwapChainResources().
Returns the number of images in the swap chain.
Note: Accessing this is necessary when providing a custom render pass and framebuffer. The framebuffer is specific to the current swapchain image and hence the application must provide multiple framebuffers.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initSwapChainResources() up until QVulkanWindowRenderer::releaseSwapChainResources().
Returns the image size of the swapchain.
This usually matches the size of the window, but may also differ in case vkGetPhysicalDeviceSurfaceCapabilitiesKHR
reports a fixed size.
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initSwapChainResources() up until QVulkanWindowRenderer::releaseSwapChainResources().
Returns the specified swap chain image view.
idx must be in the range [0, swapChainImageCount() - 1].
Note: Calling this function is only valid from the invocation of QVulkanWindowRenderer::initSwapChainResources() up until QVulkanWindowRenderer::releaseSwapChainResources().
This variable holds a constant value that is always equal to or greater than the maximum value of concurrentFrameCount().