Add 10/12 bit YUV To YUV functions

The following functions (and their 12 bit variant) are added:

planar, 10->10:
 I410ToI010, I210ToI010

planar, 10->8:
 I410ToI444, I210ToI422

planar<->biplanar, 10->10:
 I010ToP010, I210ToP210, I410ToP410
 P010ToI010, P210ToI210, P410ToI410

R=fbarchard@chromium.org

Change-Id: I9aa2bafa0d6a6e1e38ce4e20cbb437e10f9b0158
Bug: libyuv:834, libyuv:873
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/2709822
Reviewed-by: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: richard winterton <rrwinterton@gmail.com>

1 files changed, 31 insertions, 15 deletions

diff --git a/source/row_common.cc b/source/row_common.cc
index eb889c83..a941c3f5 100644
--- a/source/row_common.cc
+++ b/source/row_common.cc
@@ -2521,27 +2521,33 @@ void MergeXRGBRow_C(const uint8_t* src_r,
   }
 }
 
-// Use scale to convert lsb formats to msb, depending how many bits there are:
-// 128 = 9 bits
-// 64 = 10 bits
-// 16 = 12 bits
-// 1 = 16 bits
+// Convert lsb formats to msb, depending on sample depth.
 void MergeUVRow_16_C(const uint16_t* src_u,
                      const uint16_t* src_v,
                      uint16_t* dst_uv,
-                     int scale,
+                     int depth,
                      int width) {
+  int shift = 16 - depth;
   int x;
-  for (x = 0; x < width - 1; x += 2) {
-    dst_uv[0] = src_u[x] * scale;
-    dst_uv[1] = src_v[x] * scale;
-    dst_uv[2] = src_u[x + 1] * scale;
-    dst_uv[3] = src_v[x + 1] * scale;
-    dst_uv += 4;
+  for (x = 0; x < width; ++x) {
+    dst_uv[0] = src_u[x] << shift;
+    dst_uv[1] = src_v[x] << shift;
+    dst_uv += 2;
   }
-  if (width & 1) {
-    dst_uv[0] = src_u[width - 1] * scale;
-    dst_uv[1] = src_v[width - 1] * scale;
+}
+
+// Convert msb formats to lsb, depending on sample depth.
+void SplitUVRow_16_C(const uint16_t* src_uv,
+                     uint16_t* dst_u,
+                     uint16_t* dst_v,
+                     int depth,
+                     int width) {
+  int shift = 16 - depth;
+  int x;
+  for (x = 0; x < width; ++x) {
+    dst_u[x] = src_uv[0] >> shift;
+    dst_v[x] = src_uv[1] >> shift;
+    src_uv += 2;
   }
 }
 
@@ -2555,6 +2561,16 @@ void MultiplyRow_16_C(const uint16_t* src_y,
   }
 }
 
+void DivideRow_16_C(const uint16_t* src_y,
+                    uint16_t* dst_y,
+                    int scale,
+                    int width) {
+  int x;
+  for (x = 0; x < width; ++x) {
+    dst_y[x] = (src_y[x] * scale) >> 16;
+  }
+}
+
 // Use scale to convert lsb formats to msb, depending how many bits there are:
 // 32768 = 9 bits
 // 16384 = 10 bits