Bug 1062066 (Part 1) - Add support for vertically flipping downscaler output. r=tn

image/Downscaler.cpp

@@ -27,6 +27,7 @@ Downscaler::Downscaler(const nsIntSize& aTargetSize)
   , mYFilter(MakeUnique<skia::ConvolutionFilter1D>())
   , mWindowCapacity(0)
   , mHasAlpha(true)
+  , mFlipVertically(false)
 {
   MOZ_ASSERT(gfxPrefs::ImageDownscaleDuringDecodeEnabled(),
              "Downscaling even though downscale-during-decode is disabled?");
@@ -57,7 +58,8 @@ Downscaler::ReleaseWindow()
 nsresult
 Downscaler::BeginFrame(const nsIntSize& aOriginalSize,
                        uint8_t* aOutputBuffer,
-                       bool aHasAlpha)
+                       bool aHasAlpha,
+                       bool aFlipVertically /* = false */)
 {
   MOZ_ASSERT(aOutputBuffer);
   MOZ_ASSERT(mTargetSize != aOriginalSize,
@@ -74,6 +76,7 @@ Downscaler::BeginFrame(const nsIntSize& aOriginalSize,
                    double(mOriginalSize.height) / mTargetSize.height);
   mOutputBuffer = aOutputBuffer;
   mHasAlpha = aHasAlpha;
+  mFlipVertically = aFlipVertically;
 
   ResetForNextProgressivePass();
   ReleaseWindow();
@@ -198,9 +201,18 @@ Downscaler::TakeInvalidRect()
   DownscalerInvalidRect invalidRect;
 
   // Compute the target size invalid rect.
-  invalidRect.mTargetSizeRect =
-    nsIntRect(0, mPrevInvalidatedLine,
+  if (mFlipVertically) {
+    // We need to flip it. This will implicitly flip the original size invalid
+    // rect, since we compute it by scaling this rect.
+    invalidRect.mTargetSizeRect =
+      IntRect(0, mTargetSize.height - mCurrentOutLine,
               mTargetSize.width, mCurrentOutLine - mPrevInvalidatedLine);
+  } else {
+    invalidRect.mTargetSizeRect =
+      IntRect(0, mPrevInvalidatedLine,
+              mTargetSize.width, mCurrentOutLine - mPrevInvalidatedLine);
+  }
+
   mPrevInvalidatedLine = mCurrentOutLine;
 
   // Compute the original size invalid rect.
@@ -225,8 +237,13 @@ Downscaler::DownscaleInputLine()
   auto filterValues =
     mYFilter->FilterForValue(mCurrentOutLine, &filterOffset, &filterLength);
 
+  int32_t currentOutLine = mFlipVertically
+                         ? mTargetSize.height - (mCurrentOutLine + 1)
+                         : mCurrentOutLine;
+  MOZ_ASSERT(currentOutLine >= 0);
+
   uint8_t* outputLine =
-    &mOutputBuffer[mCurrentOutLine * mTargetSize.width * sizeof(uint32_t)];
+    &mOutputBuffer[currentOutLine * mTargetSize.width * sizeof(uint32_t)];
   skia::ConvolveVertically(static_cast<const FilterValue*>(filterValues),
                            filterLength, mWindow.get(), mXFilter->num_values(),
                            outputLine, mHasAlpha, supports_sse2());

image/Downscaler.h

@@ -70,10 +70,14 @@ public:
    *                      decode.
    * @param aHasAlpha Whether or not this frame has an alpha channel.
    *                  Performance is a little better if it doesn't have one.
+   * @param aFlipVertically If true, output rows will be written to the output
+   *                        buffer in reverse order vertically, which matches
+   *                        the way they are stored in some image formats.
    */
   nsresult BeginFrame(const nsIntSize& aOriginalSize,
                       uint8_t* aOutputBuffer,
-                      bool aHasAlpha);
+                      bool aHasAlpha,
+                      bool aFlipVertically = false);
 
   /// Retrieves the buffer into which the Decoder should write each row.
   uint8_t* RowBuffer() { return mRowBuffer.get(); }
@@ -117,7 +121,8 @@ private:
   int32_t mCurrentOutLine;
   int32_t mCurrentInLine;
 
-  bool mHasAlpha;
+  bool mHasAlpha : 1;
+  bool mFlipVertically : 1;
 };
 
 #else
@@ -139,7 +144,7 @@ public:
   const nsIntSize& TargetSize() const { return nsIntSize(); }
   const gfxSize& Scale() const { return gfxSize(1.0, 1.0); }
 
-  nsresult BeginFrame(const nsIntSize&, uint8_t*, bool)
+  nsresult BeginFrame(const nsIntSize&, uint8_t*, bool, bool = false)
   {
     return NS_ERROR_FAILURE;
   }