DirectX and XAML interop (Windows Runtime apps using DirectX with C++)

Artículo
01/20/2016

[ This article is for Windows 8.x and Windows Phone 8.x developers writing Windows Runtime apps. If you’re developing for Windows 10, see the latest documentation ]

With the release of Windows 8, you can use Extensible Application Markup Language (XAML) and Microsoft DirectX together in your Windows Store app. The combination of XAML and DirectX lets you build flexible user interface frameworks that interop with your DirectX-rendered content, and is particularly useful for graphics-intensive apps. This topic explains the structure of a Windows Store app that uses DirectX and identifies the important types to use when building your Windows Store app to work with DirectX.

Note DirectX APIs are not defined as Windows Runtime types, so you typically use Visual C++ component extensions (C++/CX) to develop XAMLWindows Store app components that interop with DirectX. Also, you can create a Windows Store app with C# and XAML that uses DirectX, if you wrap the DirectX calls in a separate Windows Runtime metadata file.

XAML and DirectX

DirectX provides two powerful libraries for 2D and 3D graphics: Direct2D and Microsoft Direct3D. Although XAML provides support for basic 2D primitives and effects, many apps, such as modeling and gaming, need more complex graphics support. For these, you can use Direct2D and Direct3D to render part or all of the graphics and use XAML for everything else.

In the XAML and DirectX interop scenario, you need to know these two concepts:

Shared surfaces are sized regions of the display, defined by XAML, that you can use DirectX to draw into indirectly, using Windows::UI::Xaml::Media::Brush types. For shared surfaces, you don't control the calls to present the swap chain(s). The updates to the shared surface are synced to the XAML framework's updates.
The swap chain itself. This provides the back buffer of the DirectX rendering pipeline, the area of memory that is presented for display after the render target is complete.

Consider what you are using DirectX for. Will it be used to composite or animate a single control that fits within the dimensions of the display window? Can the composited surface be occluded by other surfaces, or the edges of the screen? Will it contain output that needs to be rendered and controlled in real-time, as in a game?

Once you've determined how you intend to use DirectX, you use one of these Windows Runtime types to incorporate DirectX rendering into your Windows Store app:

If you want to compose a static image, or draw a complex image on event-driven intervals, draw to a shared surface withWindows::UI::Xaml::Media::Imaging::SurfaceImageSource. This type handles a sized DirectX drawing surface. Typically, you use this type when composing an image or texture as a bitmap for display in a document or UI element. It doesn't work well for real-time interactivity, such as a high-performance game. That's because updates to a SurfaceImageSource object are synced to XAML user interface updates, and that can introduce latency into the visual feedback you provide to the user, like a fluctuating frame rate or a perceived poor response to real-time input. Updates are still quick enough for dynamic controls or data simulations, though!

SurfaceImageSource graphics objects can be composited with other XAML UI elements. You can transform or project them , and the XAML framework respects any opacity or z-index values.
If the image is larger than the provided screen real estate, and can be panned or zoomed by the user, use Windows::UI::Xaml::Media::Imaging::VirtualSurfaceImageSource. This type handles a sized DirectX drawing surface that is larger than the screen. Like SurfaceImageSource, you use this when composing a complex image or control dynamically. And, also like SurfaceImageSource it doesn't work well for high-performance games. Some examples of XAML elements that could use a VirtualSurfaceImageSource are map controls, or a large, image-dense document viewer.
If you are using DirectX to present graphics updated in real-time, or in a situation where the updates must come on regular low-latency intervals, use the SwapChainPanel class, so you can refresh the graphics without syncing to the XAML framework refresh timer. This type enables you to access the graphics device's swap chain (IDXGISwapChain1) directly and layer XAML atop the render target. This type works great for games and other full-screen DirectX apps that require a XAML-based user interface. You must know DirectX well to use this approach, including the Microsoft DirectX Graphics Infrastructure (DXGI), Direct2D, and Direct3D technologies. For more info, see Programming Guide for Direct3D 11.

SurfaceImageSource

SurfaceImageSource provides DirectX shared surfaces to draw into and then composes the bits into app content.

Here is the basic process for creating and updating a SurfaceImageSource object in the code behind:

Define the size of the shared surface by passing the height and width to the SurfaceImageSource constructor. You can also indicate whether the surface needs alpha (opacity) support.

For example:

SurfaceImageSource^ surfaceImageSource = ref new SurfaceImageSource(400, 300);
Get a pointer to ISurfaceImageSourceNative. Cast the SurfaceImageSource object as IInspectable (or IUnknown), and call QueryInterface on it to get the underlying ISurfaceImageSourceNative implementation. You use the methods defined on this implementation to set the device and run the draw operations.
```
Microsoft::WRL::ComPtr<ISurfaceImageSourceNative> m_sisNative;
// ...
IInspectable* sisInspectable = (IInspectable*) reinterpret_cast<IInspectable*>(surfaceImageSource);
sisInspectable->QueryInterface(__uuidof(ISurfaceImageSourceNative), (void **)&m_sisNative);
```

Set the DXGI device by first calling D3D11CreateDevice and then passing the device and context to ISurfaceImageSourceNative::SetDevice. For example:

Microsoft::WRL::ComPtr<ID3D11Device>              m_d3dDevice;
Microsoft::WRL::ComPtr<ID3D11DeviceContext>           m_d3dContext;
// ...
D3D11CreateDevice(
        NULL,
        D3D_DRIVER_TYPE_HARDWARE,
        NULL,
        flags,
        featureLevels,
        ARRAYSIZE(featureLevels),
        D3D11_SDK_VERSION,
        &m_d3dDevice,
        NULL,
        &m_d3dContext
        )
    );  
Microsoft::WRL::ComPtr<IDXGIDevice> dxgiDevice;
m_d3dDevice.As(&dxgiDevice);
// ...
m_sisNative->SetDevice(dxgiDevice.Get());

Provide a pointer to IDXGISurface object to ISurfaceImageSourceNative::BeginDraw, and draw into that surface using DirectX. Only the area specified for update in the updateRect parameter is drawn.

Note You can only have one outstanding BeginDraw operation active at a time per IDXGIDevice.

This method returns the point (x,y) offset of the updated target rectangle in the offset parameter. You use this offset to determine where to draw into inside the IDXGISurface.
```
ComPtr<IDXGISurface> surface;

HRESULT beginDrawHR = m_sisNative->BeginDraw(updateRect, &surface, &offset);
if (beginDrawHR == DXGI_ERROR_DEVICE_REMOVED || beginDrawHR == DXGI_ERROR_DEVICE_RESET)
{
          // device changed
}
else
{
    // draw to IDXGISurface (the surface paramater)
}
```

Call ISurfaceImageSourceNative::EndDraw to complete the bitmap. Pass this bitmap to an ImageBrush.

m_sisNative->EndDraw();
// ...
// The SurfaceImageSource object's underlying ISurfaceImageSourceNative object contains the completed bitmap.
ImageBrush^ brush = ref new ImageBrush();
brush->ImageSource = surfaceImageSource;

Use the ImageBrush to draw the bitmap.

Note Calling SurfaceImageSource::SetSource (inherited from IBitmapSource::SetSource) currently throws an exception. Do not call it from your SurfaceImageSource object.

VirtualSurfaceImageSource

VirtualSurfaceImageSource extends SurfaceImageSource when the content is potentially larger than what can fit on screen and so the content must be virtualized to render optimally.

VirtualSurfaceImageSource differs from SurfaceImageSource in that it uses a callback, IVirtualSurfaceImageSourceCallbacksNative::UpdatesNeeded, that you implement to update regions of the surface as they become visible on the screen. You do not need to clear regions that are hidden, as the XAML framework takes care of that for you.

Here is basic process for creating and updating a VirtualSurfaceImageSource object in the codebehind:

Create an instance of VirtualSurfaceImageSource with the size you want. For example:

VirtualSurfaceImageSource^ virtualSIS = ref new VirtualSurfaceImageSource(2000, 2000);
Get a pointer to IVirtualSurfaceImageSourceNative. Cast the VirtualSurfaceImageSource object as IInspectable or IUnknown, and call QueryInterface on it to get the underlying IVirtualSurfaceImageSourceNative implementation. You use the methods defined on this implementation to set the device and run the draw operations.
```
Microsoft::WRL::ComPtr<IVirtualSurfaceImageSourceNative>  m_vsisNative;
// ...
IInspectable* vsisInspectable = (IInspectable*) reinterpret_cast<IInspectable*>(virtualSIS);
vsisInspectable->QueryInterface(__uuidof(IVirtualSurfaceImageSourceNative), (void **)&m_vsisNative);
```

Set the DXGI device by calling IVirtualSurfaceImageSourceNative::SetDevice. For example:

Microsoft::WRL::ComPtr<ID3D11Device>              m_d3dDevice;
Microsoft::WRL::ComPtr<ID3D11DeviceContext>           m_d3dContext;
// ...
D3D11CreateDevice(
        NULL,
        D3D_DRIVER_TYPE_HARDWARE,
        NULL,
        flags,
        featureLevels,
        ARRAYSIZE(featureLevels),
        D3D11_SDK_VERSION,
        &m_d3dDevice,
        NULL,
        &m_d3dContext
        )
    );  
Microsoft::WRL::ComPtr<IDXGIDevice> dxgiDevice;
m_d3dDevice.As(&dxgiDevice);
// ...
m_vsisNative->SetDevice(dxgiDevice.Get());

Call IVirtualSurfaceImageSourceNative::RegisterForUpdatesNeeded, passing in a reference to your implementation of IVirtualSurfaceUpdatesCallbackNative.
```
class MyContentImageSource : public IVirtualSurfaceUpdatesCallbackNative
{
// ...
  private:
     virtual HRESULT STDMETHODCALLTYPE UpdatesNeeded() override;
}

// ...

HRESULT STDMETHODCALLTYPE MyContentImageSource::UpdatesNeeded()
{
  // .. perform drawing here ...
}
void MyContentImageSource::Initialize()
{
  // ...
  m_vsisNative->RegisterForUpdatesNeeded(this);
  // ...
}
```
The framework calls your implementation of IVirtualSurfaceUpdatesCallbackNative::UpdatesNeeded when a region of the VirtualSurfaceImageSource needs to be updated.

This can happen either when the framework determines the region needs to be drawn (such as when the user pans or zooms the view of the surface), or after the app has called IVirtualSurfaceImageSourceNative::Invalidate on that region.

In IVirtualSurfaceImageSourceNative::UpdatesNeeded, use the IVirtualSurfaceImageSourceNative::GetUpdateRectCount and IVirtualSurfaceImageSourceNative::GetUpdateRects methods to determine which region(s) of the surface must be drawn.



HRESULT STDMETHODCALLTYPE MyContentImageSource::UpdatesNeeded()
{
    HRESULT hr = S_OK;

    try
    {
        ULONG drawingBoundsCount = 0;  

              m_vsisNative->GetUpdateRectCount(&drawingBoundsCount);
        std::unique_ptr<RECT[]> drawingBounds(new RECT[drawingBoundsCount]);
        m_vsisNative->GetUpdateRects(drawingBounds.get(), drawingBoundsCount);

        for (ULONG i = 0; i < drawingBoundsCount; ++i)
        {
            // Drawing code here ...
        }
    }
    catch (Platform::Exception^ exception)
    {
        hr = exception->HResult;
    }

    return hr;
}

Lastly, for each region that must be updated:
1. Provide a pointer to IDXGISurface object to IVirtualSurfaceImageSourceNative::BeginDraw, and draw into that surface using DirectX. Only the area specified for update in the updateRect parameter will be drawn.
  
  As with IlSurfaceImageSourceNative::BeginDraw, this method returns the point (x,y) offset of the updated target rectangle in the offset parameter. You use this offset to determine where to draw into inside the IDXGISurface.
  
  Note You can only have one outstanding BeginDraw operation active at a time per IDXGIDevice.
```
ComPtr<IDXGISurface> bigSurface;

HRESULT beginDrawHR = m_vsisNative->BeginDraw(updateRect, &bigSurface, &offset);
if (beginDrawHR == DXGI_ERROR_DEVICE_REMOVED || beginDrawHR == DXGI_ERROR_DEVICE_RESET)
{
          // device changed
}
else
{
    // draw to IDXGISurface
}
```
2. Draw the specific content to that region, but constrain your drawing to the bounded regions for better performance.
3. Call IVirtualSurfaceImageSourceNative::EndDraw. The result is a bitmap.

SwapChainPanel and gaming

SwapChainPanel is the Windows Runtime type designed to support high-performance graphics and gaming, where you manage the swap chain directly. In this case, you create your own DirectX swap chain and manage the presentation of your rendered content. You can then add XAML elements to the SwapChainPanel object, such as menus, heads-up displays, and other UI overlays.

To ensure good performance, there are certain limitations to the SwapChainPanel type:

There are no more than 4 SwapChainPanel instances per app.
The Opacity, RenderTransform, Projection, and Clip properties inherited by SwapChainPanel are not supported.
You should set the DirectX swap chain's height and width (in DXGI_SWAP_CHAIN_DESC1) to the current dimensions of the app window. If you don't, the display content will be scaled (using DXGI_SCALING_STRETCH) to fit.
You must set the DirectX swap chain's scaling mode (in DXGI_SWAP_CHAIN_DESC1) to DXGI_SCALING_STRETCH.
You can't set the DirectX swap chain's alpha mode (in DXGI_SWAP_CHAIN_DESC1) to DXGI_ALPHA_MODE_PREMULTIPLIED.
You must create the DirectX swap chain by calling IDXGIFactory2::CreateSwapChainForComposition.

You update the SwapChainPanel based on the needs of your app, and not the updates of the XAML framework. If you need to synchronize the updates of SwapChainPanel to those of the XAML framework, register for the Windows::UI::Xaml::Media::CompositionTarget::Rendering event. Otherwise, you must consider any cross-thread issues if you try to update the XAML elements from a different thread than the one updating the SwapChainPanel.

There are also a few general best practices to follow designing your app to use SwapChainPanel.

SwapChainPanel inherits from Windows::UI::Xaml::Controls::Grid, and supports similar layout behavior. Familiarize yourself with the Grid type and its properties.
After a DirectX swap chain has been set, all input events that are fired for SwapChainPanel work the same as they do for any other XAML element. You don't set a background brush for SwapChainPanel, and you don't need to handle input events from the app's CoreWindow object directly as you do in DirectX apps that don't use SwapChainPanel.
• All content of the visual XAML element tree under a direct child of a SwapChainPanel is clipped to the layout size of the SwapChainPanel object’s immediate child. Any content that is transformed outside these layout bounds won't be rendered. Therefore, place any XAML content that you animate with a XAMLStoryboard in the visual tree under an element whose layout bounds are large enough to contain the full range of the animation.
Limit the number of immediate visual XAML elements under a SwapChainPanel. If possible, group elements that are in close proximity under a common parent. But there is a performance tradeoff between the number of immediate visual children and the size of the children: too many or unnecessarily large XAML elements can impact overall performance. Likewise, don't create a single full-screen child XAML element for your app's SwapChainPanel because this increases overdraw in the app and decreases performance. As a rule, create no more than 8 immediate XAML visual children for your app's SwapChainPanel, and each element must have a layout size only as large as necessary to contain the element's visual content. However, you can make the visual tree of elements under a child element of the SwapChainPanel sufficiently complex without decreasing performance too badly.

Note In general, your DirectX apps should create swap chains in landscape orientation, and equal to the display window size (which is usually the native screen resolution in most Windows Store games). This ensures that your app uses the optimal swap chain implementation when it doesn't have any visible XAML overlay. If the app is rotated to portrait mode, your app should call IDXGISwapChain1::SetRotation on the existing swap chain, apply a transform to the content if needed, and then call SetSwapChain again on the same swap chain. Similarly, your app should call SetSwapChain again on the same swap chain whenever the swap chain is resized by calling IDXGISwapChain::ResizeBuffers.

Here is basic process for creating and updating a SwapChainPanel object in the code behind:

Get an instance of a swap chain panel for your app. The instances are indicated in your XAML with the <SwapChainPanel> tag.

Windows::UI::Xaml::Controls::SwapChainPanel^ swapChainPanel;

Here is an example <SwapChainPanel> tag.

<SwapChainPanel x:Name="swapChainPanel">
    <SwapChainPanel.ColumnDefinitions>
        <ColumnDefinition Width="300*"/>
        <ColumnDefinition Width="1069*"/>
    </SwapChainPanel.ColumnDefinitions>
…
</SwapChainPanel>

Get a pointer to ISwapChainPanelNative. Cast the SwapChainPanel object as IInspectable (or IUnknown), and call QueryInterface on it to get the underlying ISwapChainPanelNative implementation.

Microsoft::WRL::ComPtr<ISwapChainPanelNative> m_swapChainNative;
// ...
IInspectable* panelInspectable = (IInspectable*) reinterpret_cast<IInspectable*>(swapChainPanel);
panelInspectable->QueryInterface(__uuidof(ISwapChainPanelNative), (void **)&m_swapChainNative);

Create the DXGI device and the swap chain, and set the swap chain to ISwapChainPanelNative by passing it to SetSwapChain.

Microsoft::WRL::ComPtr<IDXGISwapChain1>               m_swapChain;    
// ...
DXGI_SWAP_CHAIN_DESC1 swapChainDesc = {0};
        swapChainDesc.Width = m_bounds.Width;
        swapChainDesc.Height = m_bounds.Height;
        swapChainDesc.Format = DXGI_FORMAT_B8G8R8A8_UNORM;           // this is the most common swapchain format
        swapChainDesc.Stereo = false; 
        swapChainDesc.SampleDesc.Count = 1;                          // don't use multi-sampling
        swapChainDesc.SampleDesc.Quality = 0;
        swapChainDesc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT;
        swapChainDesc.BufferCount = 2;
        swapChainDesc.Scaling = DXGI_SCALING_STRETCH;
        swapChainDesc.SwapEffect = DXGI_SWAP_EFFECT_FLIP_SEQUENTIAL; // we recommend using this swap effect for all applications
        swapChainDesc.Flags = 0;

// QI for DXGI device
Microsoft::WRL::ComPtr<IDXGIDevice> dxgiDevice;
m_d3dDevice.As(&dxgiDevice);

// get the DXGI adapter
Microsoft::WRL::ComPtr<IDXGIAdapter> dxgiAdapter;
dxgiDevice->GetAdapter(&dxgiAdapter);

// get the DXGI factory
Microsoft::WRL::ComPtr<IDXGIFactory2> dxgiFactory;
dxgiAdapter->GetParent(__uuidof(IDXGIFactory2), &dxgiFactory);
// create swap chain by calling CreateSwapChainForComposition
dxgiFactory->CreateSwapChainForComposition(
            m_d3dDevice.Get(),
            &swapChainDesc,
            nullptr,        // allow on any display 
            &m_swapChain
            );

m_swapChainNative->SetSwapChain(m_swapChain.Get());

Draw to the DirectX swap chain, and present it to display the contents.
```
HRESULT hr = m_swapChain->Present(1, 0);
```
The XAML elements are refreshed when the Windows Runtime layout/render logic signals an update.