Add support for odd height YUVA alpha blending.

source/planar_functions.cc

 /*
  *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS. All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
@@ skipping 633 matching lines @@
               const uint8* src_v0, int src_stride_v0,
               const uint8* src_y1, int src_stride_y1,
               const uint8* src_u1, int src_stride_u1,
               const uint8* src_v1, int src_stride_v1,
               const uint8* alpha, int alpha_stride,
               uint8* dst_y, int dst_stride_y,
               uint8* dst_u, int dst_stride_u,
               uint8* dst_v, int dst_stride_v,
               int width, int height) {
   int y;
+  // Half width/height for UV.
+  int halfwidth = (width + 1) >> 1;
   void (*BlendPlaneRow)(const uint8* src0, const uint8* src1,
       const uint8* alpha, uint8* dst, int width) = BlendPlaneRow_C;
   void (*ScaleRowDown2)(const uint8* src_ptr, ptrdiff_t src_stride,
                         uint8* dst_ptr, int dst_width) = ScaleRowDown2Box_C;
   if (!src_y0 || !src_u0 || !src_v0 || !src_y1 || !src_u1 || !src_v1 ||
       !alpha || !dst_y || !dst_u || !dst_v || width <= 0 || height == 0) {
     return -1;
   }
 
   // Negative height means invert the image.
   if (height < 0) {
     height = -height;
     dst_y = dst_y + (height - 1) * dst_stride_y;
     dst_stride_y = -dst_stride_y;
   }
 
   // Blend Y plane.
   BlendPlane(src_y0, src_stride_y0,
              src_y1, src_stride_y1,
              alpha, alpha_stride,
              dst_y, dst_stride_y,
              width, height);
 
-  // Half width/height for UV.
-  width = (width + 1) >> 1;
-  height = (height + 1) >> 1;
-
 #if defined(HAS_BLENDPLANEROW_SSSE3)
   if (TestCpuFlag(kCpuHasSSSE3)) {
 // TODO(fbarchard): Implement any versions for odd width.
 //  BlendPlaneRow = BlendPlaneRow_Any_SSSE3;
-    if (IS_ALIGNED(width, 8)) {
+    if (IS_ALIGNED(halfwidth, 8)) {
       BlendPlaneRow = BlendPlaneRow_SSSE3;
     }
   }
 #endif
 #if defined(HAS_BLENDPLANEROW_AVX2)
   if (TestCpuFlag(kCpuHasAVX2)) {
 //  BlendPlaneRow = BlendPlaneRow_Any_AVX2;
-    if (IS_ALIGNED(width, 16)) {
+    if (IS_ALIGNED(halfwidth, 16)) {
       BlendPlaneRow = BlendPlaneRow_AVX2;
     }
   }
 #endif
 #if defined(HAS_SCALEROWDOWN2_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ScaleRowDown2 = ScaleRowDown2Box_Any_NEON;
-    if (IS_ALIGNED(width, 16)) {
+    if (IS_ALIGNED(halfwidth, 16)) {
       ScaleRowDown2 = ScaleRowDown2Box_NEON;
     }
   }
 #endif
 #if defined(HAS_SCALEROWDOWN2_SSE2)
   if (TestCpuFlag(kCpuHasSSE2)) {
     ScaleRowDown2 = ScaleRowDown2Box_Any_SSE2;
-    if (IS_ALIGNED(width, 16)) {
+    if (IS_ALIGNED(halfwidth, 16)) {
       ScaleRowDown2 = ScaleRowDown2Box_SSE2;
     }
   }
 #endif
 #if defined(HAS_SCALEROWDOWN2_AVX2)
   if (TestCpuFlag(kCpuHasAVX2)) {
     ScaleRowDown2 = ScaleRowDown2Box_Any_AVX2;
-    if (IS_ALIGNED(width, 32)) {
+    if (IS_ALIGNED(halfwidth, 32)) {
       ScaleRowDown2 = ScaleRowDown2Box_AVX2;
     }
   }
 #endif
 
   // Row buffer for intermediate alpha pixels.
-  align_buffer_64(halfalpha, width);
-  for (y = 0; y < height; ++y) {
+  align_buffer_64(halfalpha, halfwidth);
+  for (y = 0; y < height; y += 2) {
+    // last row of odd height image use 1 row of alpha instead of 2.
+    if (y == (height - 1)) {
+      alpha_stride = 0;
+    }
     // Subsample 2 rows of UV to half width and half height.
-    ScaleRowDown2(alpha, alpha_stride, halfalpha, width);
+    ScaleRowDown2(alpha, alpha_stride, halfalpha, halfwidth);
     alpha += alpha_stride * 2;
-    BlendPlaneRow(src_u0, src_u1, halfalpha, dst_u, width);
-    BlendPlaneRow(src_v0, src_v1, halfalpha, dst_v, width);
+    BlendPlaneRow(src_u0, src_u1, halfalpha, dst_u, halfwidth);
+    BlendPlaneRow(src_v0, src_v1, halfalpha, dst_v, halfwidth);
     src_u0 += src_stride_u0;
     src_u1 += src_stride_u1;
     dst_u += dst_stride_u;
     src_v0 += src_stride_v0;
     src_v1 += src_stride_v1;
     dst_v += dst_stride_v;
   }
   free_aligned_buffer_64(halfalpha);
   return 0;
 }
@@ skipping 1884 matching lines @@
     }
     free_aligned_buffer_64(rows);
   }
   return 0;
 }
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif