libyuv roll to r1652 for new I422ToRGB565 function

Bug: none
Test: I422ToRGB565 unittest
Change-Id: I8b4418badf725d3ecb6735effcb782976b27979b

3 files changed, 102 insertions, 33 deletions

diff --git a/files/source/convert_from.cc b/files/source/convert_from.cc
index ab5afc06..d623731d 100644
--- a/files/source/convert_from.cc
+++ b/files/source/convert_from.cc
@@ -890,6 +890,75 @@ int I420ToRGB565(const uint8* src_y,
   return 0;
 }
 
+// Convert I422 to RGB565.
+LIBYUV_API
+int I422ToRGB565(const uint8* src_y,
+                 int src_stride_y,
+                 const uint8* src_u,
+                 int src_stride_u,
+                 const uint8* src_v,
+                 int src_stride_v,
+                 uint8* dst_rgb565,
+                 int dst_stride_rgb565,
+                 int width,
+                 int height) {
+  int y;
+  void (*I422ToRGB565Row)(const uint8* y_buf, const uint8* u_buf,
+                          const uint8* v_buf, uint8* rgb_buf,
+                          const struct YuvConstants* yuvconstants, int width) =
+      I422ToRGB565Row_C;
+  if (!src_y || !src_u || !src_v || !dst_rgb565 || width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_rgb565 = dst_rgb565 + (height - 1) * dst_stride_rgb565;
+    dst_stride_rgb565 = -dst_stride_rgb565;
+  }
+#if defined(HAS_I422TORGB565ROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB565Row = I422ToRGB565Row_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB565Row = I422ToRGB565Row_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB565Row = I422ToRGB565Row_NEON;
+    }
+  }
+#endif
+#if defined(HAS_I422TORGB565ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB565Row = I422ToRGB565Row_MSA;
+    }
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    I422ToRGB565Row(src_y, src_u, src_v, dst_rgb565, &kYuvI601Constants, width);
+    dst_rgb565 += dst_stride_rgb565;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
 // Ordered 8x8 dither for 888 to 565.  Values from 0 to 7.
 static const uint8 kDither565_4x4[16] = {
     0, 4, 1, 5, 6, 2, 7, 3, 1, 5, 0, 4, 7, 3, 6, 2,
diff --git a/files/source/rotate_win.cc b/files/source/rotate_win.cc
index 044048d3..93a5c28a 100644
--- a/files/source/rotate_win.cc
+++ b/files/source/rotate_win.cc
@@ -137,9 +137,9 @@ __declspec(naked) void TransposeUVWx8_SSE2(const uint8* src,
     mov       ecx, [ecx + 16 + 28]  // w
 
     align      4
- convertloop:
-                     // Read in the data from the source pointer.
-        // First round of bit swap.
+    // Read in the data from the source pointer.
+    // First round of bit swap.
+  convertloop:
     movdqu    xmm0, [eax]
     movdqu    xmm1, [eax + edi]
     lea       eax, [eax + 2 * edi]
@@ -166,7 +166,7 @@ __declspec(naked) void TransposeUVWx8_SSE2(const uint8* src,
     lea       eax, [eax + 2 * edi]
     movdqu    [esp], xmm5  // backup xmm5
     neg       edi
-    movdqa    xmm5, xmm6             // use xmm5 as temp register.
+    movdqa    xmm5, xmm6  // use xmm5 as temp register.
     punpcklbw xmm6, xmm7
     punpckhbw xmm5, xmm7
     movdqa    xmm7, xmm5
@@ -187,10 +187,11 @@ __declspec(naked) void TransposeUVWx8_SSE2(const uint8* src,
     movdqa    xmm6, xmm5
     movdqu    xmm5, [esp]  // restore xmm5
     movdqu    [esp], xmm6  // backup xmm6
-    movdqa    xmm6, xmm5             // use xmm6 as temp register.
+    movdqa    xmm6, xmm5  // use xmm6 as temp register.
     punpcklwd xmm5, xmm7
     punpckhwd xmm6, xmm7
     movdqa    xmm7, xmm6
+
     // Third round of bit swap.
     // Write to the destination pointer.
     movdqa    xmm6, xmm0
diff --git a/files/source/row_any.cc b/files/source/row_any.cc
index af5d1fbc..1092a9c0 100644
--- a/files/source/row_any.cc
+++ b/files/source/row_any.cc
@@ -158,13 +158,13 @@ ANY31C(I422ToRGBARow_Any_AVX2, I422ToRGBARow_AVX2, 1, 0, 4, 15)
 ANY31C(I444ToARGBRow_Any_AVX2, I444ToARGBRow_AVX2, 0, 0, 4, 15)
 #endif
 #ifdef HAS_I422TOARGB4444ROW_AVX2
-ANY31C(I422ToARGB4444Row_Any_AVX2, I422ToARGB4444Row_AVX2, 1, 0, 2, 7)
+ANY31C(I422ToARGB4444Row_Any_AVX2, I422ToARGB4444Row_AVX2, 1, 0, 2, 15)
 #endif
 #ifdef HAS_I422TOARGB1555ROW_AVX2
-ANY31C(I422ToARGB1555Row_Any_AVX2, I422ToARGB1555Row_AVX2, 1, 0, 2, 7)
+ANY31C(I422ToARGB1555Row_Any_AVX2, I422ToARGB1555Row_AVX2, 1, 0, 2, 15)
 #endif
 #ifdef HAS_I422TORGB565ROW_AVX2
-ANY31C(I422ToRGB565Row_Any_AVX2, I422ToRGB565Row_AVX2, 1, 0, 2, 7)
+ANY31C(I422ToRGB565Row_Any_AVX2, I422ToRGB565Row_AVX2, 1, 0, 2, 15)
 #endif
 #ifdef HAS_I422TOARGBROW_NEON
 ANY31C(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, 0, 0, 4, 7)
@@ -654,18 +654,18 @@ ANY11B(ARGBCopyYToAlphaRow_Any_SSE2, ARGBCopyYToAlphaRow_SSE2, 0, 1, 4, 7)
 #undef ANY11B
 
 // Any 1 to 1 with parameter.
-#define ANY11P(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK)                         \
-  void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, T shuffler, int width) { \
-    SIMD_ALIGNED(uint8 temp[64 * 2]);                                         \
-    memset(temp, 0, 64); /* for msan */                                       \
-    int r = width & MASK;                                                     \
-    int n = width & ~MASK;                                                    \
-    if (n > 0) {                                                              \
-      ANY_SIMD(src_ptr, dst_ptr, shuffler, n);                                \
-    }                                                                         \
-    memcpy(temp, src_ptr + n * SBPP, r * SBPP);                               \
-    ANY_SIMD(temp, temp + 64, shuffler, MASK + 1);                            \
-    memcpy(dst_ptr + n * BPP, temp + 64, r * BPP);                            \
+#define ANY11P(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK)                      \
+  void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, T param, int width) { \
+    SIMD_ALIGNED(uint8 temp[64 * 2]);                                      \
+    memset(temp, 0, 64); /* for msan */                                    \
+    int r = width & MASK;                                                  \
+    int n = width & ~MASK;                                                 \
+    if (n > 0) {                                                           \
+      ANY_SIMD(src_ptr, dst_ptr, param, n);                                \
+    }                                                                      \
+    memcpy(temp, src_ptr + n * SBPP, r * SBPP);                            \
+    ANY_SIMD(temp, temp + 64, param, MASK + 1);                            \
+    memcpy(dst_ptr + n * BPP, temp + 64, r * BPP);                         \
   }
 
 #if defined(HAS_ARGBTORGB565DITHERROW_SSE2)
@@ -718,19 +718,18 @@ ANY11P(ARGBShuffleRow_Any_MSA, ARGBShuffleRow_MSA, const uint8*, 4, 4, 7)
 #undef ANY11P
 
 // Any 1 to 1 with parameter and shorts.  BPP measures in shorts.
-#define ANY11P16(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK)            \
-  void NAMEANY(const uint16* src_ptr, uint16* dst_ptr, T shuffler, \
-               int width) {                                        \
-    SIMD_ALIGNED(uint16 temp[16 * 2]);                             \
-    memset(temp, 0, 32); /* for msan */                            \
-    int r = width & MASK;                                          \
-    int n = width & ~MASK;                                         \
-    if (n > 0) {                                                   \
-      ANY_SIMD(src_ptr, dst_ptr, shuffler, n);                     \
-    }                                                              \
-    memcpy(temp, src_ptr + n, r * SBPP);                           \
-    ANY_SIMD(temp, temp + 16, shuffler, MASK + 1);                 \
-    memcpy(dst_ptr + n, temp + 16, r * BPP);                       \
+#define ANY11P16(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK)                      \
+  void NAMEANY(const uint16* src_ptr, uint16* dst_ptr, T param, int width) { \
+    SIMD_ALIGNED(uint16 temp[16 * 2]);                                       \
+    memset(temp, 0, 32); /* for msan */                                      \
+    int r = width & MASK;                                                    \
+    int n = width & ~MASK;                                                   \
+    if (n > 0) {                                                             \
+      ANY_SIMD(src_ptr, dst_ptr, param, n);                                  \
+    }                                                                        \
+    memcpy(temp, src_ptr + n, r * SBPP);                                     \
+    ANY_SIMD(temp, temp + 16, param, MASK + 1);                              \
+    memcpy(dst_ptr + n, temp + 16, r * BPP);                                 \
   }
 
 #ifdef HAS_HALFFLOATROW_SSE2