Add SplitUVPlanes and MergeUVPlanes

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 13 matching lines @@
 extern "C" {
 #endif
 
 // Copy a plane of data
 LIBYUV_API
 void CopyPlane(const uint8* src_y, int src_stride_y,
                uint8* dst_y, int dst_stride_y,
                int width, int height) {
   int y;
   void (*CopyRow)(const uint8* src, uint8* dst, int width) = CopyRow_C;
+  // Negative height means invert the image.

[fbarchard1, 2016/08/24 23:04:11]

this change isnt needed? remove

[fbarchard1, 2016/08/24 23:11:50]

Acknowledged.  Currently needed in this CL and consistent with SplitUVPlane. 
But CopyPlane_16 does not support negative height, so added todo to reconsider
this.

+  if (height < 0) {
+    height = -height;
+    dst_y = dst_y + (height - 1) * dst_stride_y;
+    dst_stride_y = -dst_stride_y;
+  }
   // Coalesce rows.
   if (src_stride_y == width &&
       dst_stride_y == width) {
     width *= height;
     height = 1;
     src_stride_y = dst_stride_y = 0;
   }
   // Nothing to do.
   if (src_y == dst_y && src_stride_y == dst_stride_y) {
     return;
@@ skipping 2431 matching lines @@
 // TODO(fbarchard): Consider if width is even Y channel can be split
 // directly. A SplitUVRow_Odd function could copy the remaining chroma.
 
 LIBYUV_API
 int YUY2ToNV12(const uint8* src_yuy2, int src_stride_yuy2,
                uint8* dst_y, int dst_stride_y,
                uint8* dst_uv, int dst_stride_uv,
                int width, int height) {
   int y;
   int halfwidth = (width + 1) >> 1;
-  void (*SplitUVRow)(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
-                     int width) = SplitUVRow_C;
   void (*InterpolateRow)(uint8* dst_ptr, const uint8* src_ptr,
                          ptrdiff_t src_stride, int dst_width,
                          int source_y_fraction) = InterpolateRow_C;
   if (!src_yuy2 ||
       !dst_y || !dst_uv ||
       width <= 0 || height == 0) {
     return -1;
   }
   // Negative height means invert the image.
   if (height < 0) {
     height = -height;
     src_yuy2 = src_yuy2 + (height - 1) * src_stride_yuy2;
     src_stride_yuy2 = -src_stride_yuy2;
   }
-#if defined(HAS_SPLITUVROW_SSE2)
-  if (TestCpuFlag(kCpuHasSSE2)) {
-    SplitUVRow = SplitUVRow_Any_SSE2;
-    if (IS_ALIGNED(width, 16)) {
-      SplitUVRow = SplitUVRow_SSE2;
-    }
-  }
-#endif
-#if defined(HAS_SPLITUVROW_AVX2)
-  if (TestCpuFlag(kCpuHasAVX2)) {
-    SplitUVRow = SplitUVRow_Any_AVX2;
-    if (IS_ALIGNED(width, 32)) {
-      SplitUVRow = SplitUVRow_AVX2;
-    }
-  }
-#endif
-#if defined(HAS_SPLITUVROW_NEON)
-  if (TestCpuFlag(kCpuHasNEON)) {
-    SplitUVRow = SplitUVRow_Any_NEON;
-    if (IS_ALIGNED(width, 16)) {
-      SplitUVRow = SplitUVRow_NEON;
-    }
-  }
-#endif
 #if defined(HAS_INTERPOLATEROW_SSSE3)
   if (TestCpuFlag(kCpuHasSSSE3)) {
     InterpolateRow = InterpolateRow_Any_SSSE3;
     if (IS_ALIGNED(width, 16)) {
       InterpolateRow = InterpolateRow_SSSE3;
     }
   }
 #endif
 #if defined(HAS_INTERPOLATEROW_AVX2)
   if (TestCpuFlag(kCpuHasAVX2)) {
@@ skipping 10 matching lines @@
       InterpolateRow = InterpolateRow_NEON;
     }
   }
 #endif
 
   {
     int awidth = halfwidth * 2;
     // row of y and 2 rows of uv
     align_buffer_64(rows, awidth * 3);
 
+    SplitUVRowFunction SplitUVRow = GetOptimizedSplitUVRowFunction(
+        src_yuy2, src_stride_yuy2,
+        rows, awidth,
+        rows, awidth,
+        awidth);
+
     for (y = 0; y < height - 1; y += 2) {
       // Split Y from UV.
       SplitUVRow(src_yuy2, rows, rows + awidth, awidth);
       memcpy(dst_y, rows, width);
       SplitUVRow(src_yuy2 + src_stride_yuy2, rows, rows + awidth * 2, awidth);
       memcpy(dst_y + dst_stride_y, rows, width);
       InterpolateRow(dst_uv, rows + awidth, awidth, awidth, 128);
       src_yuy2 += src_stride_yuy2 * 2;
       dst_y += dst_stride_y * 2;
       dst_uv += dst_stride_uv;
     }
     if (height & 1) {
       // Split Y from UV.
-      SplitUVRow(src_yuy2, rows, dst_uv, awidth);
+      SplitUVRow(src_yuy2, rows, rows + awidth, awidth);
       memcpy(dst_y, rows, width);
+      memcpy(dst_uv, rows + awidth, awidth);

[fbarchard1, 2016/08/24 23:04:11]

this change isnt needed?  remove.
if this change is needed, do it in different CL

     }
     free_aligned_buffer_64(rows);
   }
   return 0;
 }
 
 LIBYUV_API
 int UYVYToNV12(const uint8* src_uyvy, int src_stride_uyvy,
                uint8* dst_y, int dst_stride_y,
                uint8* dst_uv, int dst_stride_uv,
                int width, int height) {
   int y;
   int halfwidth = (width + 1) >> 1;
-  void (*SplitUVRow)(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
-                     int width) = SplitUVRow_C;
   void (*InterpolateRow)(uint8* dst_ptr, const uint8* src_ptr,
                          ptrdiff_t src_stride, int dst_width,
                          int source_y_fraction) = InterpolateRow_C;
   if (!src_uyvy ||
       !dst_y || !dst_uv ||
       width <= 0 || height == 0) {
     return -1;
   }
   // Negative height means invert the image.
   if (height < 0) {
     height = -height;
     src_uyvy = src_uyvy + (height - 1) * src_stride_uyvy;
     src_stride_uyvy = -src_stride_uyvy;
   }
-#if defined(HAS_SPLITUVROW_SSE2)
-  if (TestCpuFlag(kCpuHasSSE2)) {
-    SplitUVRow = SplitUVRow_Any_SSE2;
-    if (IS_ALIGNED(width, 16)) {
-      SplitUVRow = SplitUVRow_SSE2;
-    }
-  }
-#endif
-#if defined(HAS_SPLITUVROW_AVX2)
-  if (TestCpuFlag(kCpuHasAVX2)) {
-    SplitUVRow = SplitUVRow_Any_AVX2;
-    if (IS_ALIGNED(width, 32)) {
-      SplitUVRow = SplitUVRow_AVX2;
-    }
-  }
-#endif
-#if defined(HAS_SPLITUVROW_NEON)
-  if (TestCpuFlag(kCpuHasNEON)) {
-    SplitUVRow = SplitUVRow_Any_NEON;
-    if (IS_ALIGNED(width, 16)) {
-      SplitUVRow = SplitUVRow_NEON;
-    }
-  }
-#endif
 #if defined(HAS_INTERPOLATEROW_SSSE3)
   if (TestCpuFlag(kCpuHasSSSE3)) {
     InterpolateRow = InterpolateRow_Any_SSSE3;
     if (IS_ALIGNED(width, 16)) {
       InterpolateRow = InterpolateRow_SSSE3;
     }
   }
 #endif
 #if defined(HAS_INTERPOLATEROW_AVX2)
   if (TestCpuFlag(kCpuHasAVX2)) {
@@ skipping 10 matching lines @@
       InterpolateRow = InterpolateRow_NEON;
     }
   }
 #endif
 
   {
     int awidth = halfwidth * 2;
     // row of y and 2 rows of uv
     align_buffer_64(rows, awidth * 3);
 
+    SplitUVRowFunction SplitUVRow = GetOptimizedSplitUVRowFunction(
+        src_uyvy, src_stride_uyvy,
+        rows, awidth,
+        rows, awidth,
+        awidth);
+
     for (y = 0; y < height - 1; y += 2) {
       // Split Y from UV.
       SplitUVRow(src_uyvy, rows + awidth, rows, awidth);
       memcpy(dst_y, rows, width);
       SplitUVRow(src_uyvy + src_stride_uyvy, rows + awidth * 2, rows, awidth);
       memcpy(dst_y + dst_stride_y, rows, width);
       InterpolateRow(dst_uv, rows + awidth, awidth, awidth, 128);
       src_uyvy += src_stride_uyvy * 2;
       dst_y += dst_stride_y * 2;
       dst_uv += dst_stride_uv;
     }
     if (height & 1) {
       // Split Y from UV.
-      SplitUVRow(src_uyvy, dst_uv, rows, awidth);
+      SplitUVRow(src_uyvy, rows + awidth, rows, awidth);
       memcpy(dst_y, rows, width);
+      memcpy(dst_uv, rows + awidth, awidth);
     }
     free_aligned_buffer_64(rows);
   }
   return 0;
 }
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif