YUY2ToNV12 using YUY2ToY and YUY2ToNVUV

- Optimized YUY2ToNV12 that reduces it from 3 steps to 2 steps
  - Was SplitUV, memcpy Y, InterpolateUV
  - Now YUY2ToY, YUY2ToNVUV
- rollback LIBYUV_UNLIMITED_DATA

3840x2160 1000 iterations:

Pixel 2 Cortex A73
Was YUY2ToNV12_Opt (6515 ms)
Now YUY2ToNV12_Opt (3350 ms)

AB7 Mediatek P35 Cortex A53
Was YUY2ToNV12_Opt (6435 ms)
Now YUY2ToNV12_Opt (3301 ms)

Skylake AVX2 x64
Was YUY2ToNV12_Opt (1872 ms)
Now YUY2ToNV12_Opt (1657 ms)

SSE2 x64
Was YUY2ToNV12_Opt (2008 ms)
Now YUY2ToNV12_Opt (1691 ms)

Windows Skylake AVX2 32 bit x86
Was YUY2ToNV12_Opt (2161 ms)
Now YUY2ToNV12_Opt (1628 ms)

Bug: libyuv:943
Change-Id: I6c2ba2ae765413426baf770b837de114f808f6d0
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/3929843
Reviewed-by: Wan-Teh Chang <wtc@google.com>
Reviewed-by: richard winterton <rrwinterton@gmail.com>
Commit-Queue: Frank Barchard <fbarchard@chromium.org>

1 files changed, 25 insertions, 6 deletions

diff --git a/source/row_common.cc b/source/row_common.cc
index a177d8bb..8bfa4b8c 100644
--- a/source/row_common.cc
+++ b/source/row_common.cc
@@ -22,9 +22,13 @@ extern "C" {
 #endif
 
 // This macro controls YUV to RGB using unsigned math to extend range of
-// YUV to RGB coefficients to 0 to 4 instead of 0 to 2 for more accuracy on B.
-// Enable this macro for backwards compatiability with limited range 0 to 2.
-// LIBYUV_LIMITED_DATA
+// YUV to RGB coefficients to 0 to 4 instead of 0 to 2 for more accuracy on B:
+// LIBYUV_UNLIMITED_DATA
+
+// Macros to enable unlimited data for each colorspace
+// LIBYUV_UNLIMITED_BT601
+// LIBYUV_UNLIMITED_BT709
+// LIBYUV_UNLIMITED_BT2020
 
 // The following macro from row_win makes the C code match the row_win code,
 // which is 7 bit fixed point for ARGBToI420:
@@ -1480,7 +1484,7 @@ void J400ToARGBRow_C(const uint8_t* src_y, uint8_t* dst_argb, int width) {
 // KR = 0.299; KB = 0.114
 
 // U and V contributions to R,G,B.
-#if !defined(LIBYUV_LIMITED_DATA)
+#if defined(LIBYUV_UNLIMITED_DATA) || defined(LIBYUV_UNLIMITED_BT601)
 #define UB 129 /* round(2.018 * 64) */
 #else
 #define UB 128 /* max(128, round(2.018 * 64)) */
@@ -1534,7 +1538,7 @@ MAKEYUVCONSTANTS(JPEG, YG, YB, UB, UG, VG, VR)
 //  KR = 0.2126, KB = 0.0722
 
 // U and V contributions to R,G,B.
-#if !defined(LIBYUV_LIMITED_DATA)
+#if defined(LIBYUV_UNLIMITED_DATA) || defined(LIBYUV_UNLIMITED_BT709)
 #define UB 135 /* round(2.112 * 64) */
 #else
 #define UB 128 /* max(128, round(2.112 * 64)) */
@@ -1588,7 +1592,7 @@ MAKEYUVCONSTANTS(F709, YG, YB, UB, UG, VG, VR)
 // KR = 0.2627; KB = 0.0593
 
 // U and V contributions to R,G,B.
-#if !defined(LIBYUV_LIMITED_DATA)
+#if defined(LIBYUV_UNLIMITED_DATA) || defined(LIBYUV_UNLIMITED_BT2020)
 #define UB 137 /* round(2.142 * 64) */
 #else
 #define UB 128 /* max(128, round(2.142 * 64)) */
@@ -3094,6 +3098,21 @@ void YUY2ToUVRow_C(const uint8_t* src_yuy2,
   }
 }
 
+// Filter 2 rows of YUY2 UV's (422) into UV (NV12).
+void YUY2ToNVUVRow_C(const uint8_t* src_yuy2,
+                     int src_stride_yuy2,
+                     uint8_t* dst_uv,
+                     int width) {
+  // Output a row of UV values, filtering 2 rows of YUY2.
+  int x;
+  for (x = 0; x < width; x += 2) {
+    dst_uv[0] = (src_yuy2[1] + src_yuy2[src_stride_yuy2 + 1] + 1) >> 1;
+    dst_uv[1] = (src_yuy2[3] + src_yuy2[src_stride_yuy2 + 3] + 1) >> 1;
+    src_yuy2 += 4;
+    dst_uv += 2;
+  }
+}
+
 // Copy row of YUY2 UV's (422) into U and V (422).
 void YUY2ToUV422Row_C(const uint8_t* src_yuy2,
                       uint8_t* dst_u,