Direct2D QPA: Draw directly to swap chain

Remove the intermediate pixmap in the backing store and draw directly to
the window swap chain. This is faster and reduces memory pressure on the
graphics card.

In case of native child widgets we need to read back the back buffer,
which incurs an extra copy in this case.

In an artificial benchmark drawing animated full screen
gradients as fast as possible this patch increases performance by 42% on
my current machine from 480fps to around 680fps, i.e. the time for
actually getting the pixels to the screen is now lower.

Change-Id: Ifbeda0e199afec03cecfe76337679a9e9d082bdd
Reviewed-by: Risto Avila <risto.avila@digia.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@digia.com>
Reviewed-by: Andrew Knight <andrew.knight@digia.com>

1 files changed, 6 insertions, 1 deletions

diff --git a/src/plugins/platforms/direct2d/qwindowsdirect2dwindow.h b/src/plugins/platforms/direct2d/qwindowsdirect2dwindow.h
index 7996904639..47c790da5d 100644
--- a/src/plugins/platforms/direct2d/qwindowsdirect2dwindow.h
+++ b/src/plugins/platforms/direct2d/qwindowsdirect2dwindow.h
@@ -56,16 +56,21 @@ public:
     QWindowsDirect2DWindow(QWindow *window, const QWindowsWindowData &data);
     ~QWindowsDirect2DWindow();
 
+    QPixmap *pixmap();
     void flush(QWindowsDirect2DBitmap *bitmap, const QRegion &region, const QPoint &offset);
+    void present();
+    void resizeSwapChain(const QSize &size);
+
+    QSharedPointer<QWindowsDirect2DBitmap> copyBackBuffer() const;
 
 private:
-    void resizeSwapChain(const QSize &size);
     void setupBitmap();
 
 private:
     Microsoft::WRL::ComPtr<IDXGISwapChain1> m_swapChain;
     Microsoft::WRL::ComPtr<ID2D1DeviceContext> m_deviceContext;
     QScopedPointer<QWindowsDirect2DBitmap> m_bitmap;
+    QScopedPointer<QPixmap> m_pixmap;
     bool m_needsFullFlush;
 };