Add MSA optimized I422AlphaToARGBRow_MSA and I422ToRGB24Row_MSA functions

R=fbarchard@google.com
BUG=libyuv:634

Performance Gain (vs C vectorized)
I422AlphaToARGBRow_MSA      : ~1.4x
I422AlphaToARGBRow_Any_MSA  : ~1.4x
I422ToRGB24Row_MSA          : ~4.8x
I422ToRGB24Row_Any_MSA      : ~4.8x

Performance Gain (vs C non-vectorized)
I422AlphaToARGBRow_MSA      : ~7.0x
I422AlphaToARGBRow_Any_MSA  : ~7.0x
I422ToRGB24Row_MSA          : ~7.9x
I422ToRGB24Row_Any_MSA      : ~7.7x

Review URL: https://codereview.chromium.org/2454433003 .

README.chromium

 Name: libyuv
 URL: http://code.google.com/p/libyuv/
-Version: 1631
+Version: 1632
 License: BSD
 License File: LICENSE
 

include/libyuv/row.h

@@ -383,6 +383,8 @@ extern "C" {
 #define HAS_ARGBTOUVROW_MSA
 #define HAS_I422TOARGBROW_MSA
 #define HAS_I422TORGBAROW_MSA
+#define HAS_I422ALPHATOARGBROW_MSA
+#define HAS_I422TORGB24ROW_MSA
 #endif
 
 #if defined(_MSC_VER) && !defined(__CLR_VER) && !defined(__clang__)
@@ -663,6 +665,19 @@ void I422ToRGBARow_MSA(const uint8* src_y,
                        uint8* dst_rgba,
                        const struct YuvConstants* yuvconstants,
                        int width);
+void I422AlphaToARGBRow_MSA(const uint8* y_buf,
+                            const uint8* u_buf,
+                            const uint8* v_buf,
+                            const uint8* a_buf,
+                            uint8* dst_argb,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
+void I422ToRGB24Row_MSA(const uint8* src_y,
+                        const uint8* src_u,
+                        const uint8* src_v,
+                        uint8* dst_rgb24,
+                        const struct YuvConstants* yuvconstants,
+                        int width);
 
 void ARGBToYRow_AVX2(const uint8* src_argb, uint8* dst_y, int width);
 void ARGBToYRow_Any_AVX2(const uint8* src_argb, uint8* dst_y, int width);
@@ -1653,6 +1668,19 @@ void I422ToRGBARow_Any_MSA(const uint8* src_y,
                            uint8* dst_argb,
                            const struct YuvConstants* yuvconstants,
                            int width);
+void I422AlphaToARGBRow_Any_MSA(const uint8* src_y,
+                                const uint8* src_u,
+                                const uint8* src_v,
+                                const uint8* src_a,
+                                uint8* dst_argb,
+                                const struct YuvConstants* yuvconstants,
+                                int width);
+void I422ToRGB24Row_Any_MSA(const uint8* src_y,
+                            const uint8* src_u,
+                            const uint8* src_v,
+                            uint8* dst_rgb24,
+                            const struct YuvConstants* yuvconstants,
+                            int width);
 
 void YUY2ToYRow_AVX2(const uint8* src_yuy2, uint8* dst_y, int width);
 void YUY2ToUVRow_AVX2(const uint8* src_yuy2, int stride_yuy2,

include/libyuv/version.h

@@ -11,6 +11,6 @@
 #ifndef INCLUDE_LIBYUV_VERSION_H_
 #define INCLUDE_LIBYUV_VERSION_H_
 
-#define LIBYUV_VERSION 1631
+#define LIBYUV_VERSION 1632
 
 #endif  // INCLUDE_LIBYUV_VERSION_H_

source/convert_argb.cc

@@ -564,6 +564,14 @@ static int I420AlphaToARGBMatrix(const uint8* src_y, int src_stride_y,
     I422AlphaToARGBRow = I422AlphaToARGBRow_DSPR2;
   }
 #endif
+#if defined(HAS_I422ALPHATOARGBROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422AlphaToARGBRow = I422AlphaToARGBRow_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422AlphaToARGBRow = I422AlphaToARGBRow_MSA;
+    }
+  }
+#endif
 #if defined(HAS_ARGBATTENUATEROW_SSSE3)
   if (TestCpuFlag(kCpuHasSSSE3)) {
     ARGBAttenuateRow = ARGBAttenuateRow_Any_SSSE3;

source/convert_from.cc

@@ -558,6 +558,14 @@ static int I420ToRGB24Matrix(const uint8* src_y, int src_stride_y,
     }
   }
 #endif
+#if defined(HAS_I422TORGB24ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToRGB24Row = I422ToRGB24Row_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      I422ToRGB24Row = I422ToRGB24Row_MSA;
+    }
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     I422ToRGB24Row(src_y, src_u, src_v, dst_rgb24, yuvconstants, width);

source/row_any.cc

@@ -53,6 +53,9 @@ ANY41C(I422AlphaToARGBRow_Any_AVX2, I422AlphaToARGBRow_AVX2, 1, 0, 4, 15)
 #ifdef HAS_I422ALPHATOARGBROW_NEON
 ANY41C(I422AlphaToARGBRow_Any_NEON, I422AlphaToARGBRow_NEON, 1, 0, 4, 7)
 #endif
+#ifdef HAS_I422ALPHATOARGBROW_MSA
+ANY41C(I422AlphaToARGBRow_Any_MSA, I422AlphaToARGBRow_MSA, 1, 0, 4, 7)
+#endif
 #undef ANY41C
 
 // Any 3 planes to 1.
@@ -168,6 +171,7 @@ ANY31C(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, 1, 0, 2, 7)
 #ifdef HAS_I422TOARGBROW_MSA
 ANY31C(I422ToARGBRow_Any_MSA, I422ToARGBRow_MSA, 1, 0, 4, 7)
 ANY31C(I422ToRGBARow_Any_MSA, I422ToRGBARow_MSA, 1, 0, 4, 7)
+ANY31C(I422ToRGB24Row_Any_MSA, I422ToRGB24Row_MSA, 1, 0, 3, 15)
 #endif
 #undef ANY31C
 

source/row_msa.cc

@@ -19,65 +19,79 @@ namespace libyuv {
 extern "C" {
 #endif
 
+// Load YUV 422 pixel data
+#define READYUV422(psrc_y, psrc_u, psrc_v, out_y, out_u, out_v) {  \
+  uint64 y_m;                                                      \
+  uint32 u_m, v_m;                                                 \
+  v4i32 zero_m = { 0 };                                            \
+  y_m = LD(psrc_y);                                                \
+  u_m = LW(psrc_u);                                                \
+  v_m = LW(psrc_v);                                                \
+  out_y = (v16u8) __msa_insert_d((v2i64) zero_m, 0, (int64) y_m);  \
+  out_u = (v16u8) __msa_insert_w(zero_m, 0, (int32) u_m);          \
+  out_v = (v16u8) __msa_insert_w(zero_m, 0, (int32) v_m);          \
+}
+
 // Convert 8 pixels of YUV 420 to RGB.
-#define I422TORGB(in0, in1, in2, ub, vr, ug, vg,               \
-                  bb, bg, br, yg, out0, out1, out2) {          \
-  v8i16 vec0_m;                                                \
-  v4i32 reg0_m, reg1_m, reg2_m, reg3_m, reg4_m;                \
-  v4i32 reg5_m, reg6_m, reg7_m, reg8_m, reg9_m;                \
-  v4i32 max_val_m = __msa_ldi_w(255);                          \
-  v8i16 zero_m = { 0 };                                        \
-                                                               \
-  in1 = (v16u8) __msa_ilvr_b((v16i8) in1, (v16i8) in1);        \
-  in2 = (v16u8) __msa_ilvr_b((v16i8) in2, (v16i8) in2);        \
-  vec0_m = (v8i16) __msa_ilvr_b((v16i8) in0, (v16i8) in0);     \
-  reg0_m = (v4i32) __msa_ilvr_h(zero_m, vec0_m);               \
-  reg1_m = (v4i32) __msa_ilvl_h(zero_m, vec0_m);               \
-  reg0_m *= vec_yg;                                            \
-  reg1_m *= vec_yg;                                            \
-  reg0_m = __msa_srai_w(reg0_m, 16);                           \
-  reg1_m = __msa_srai_w(reg1_m, 16);                           \
-  reg4_m = reg0_m + br;                                        \
-  reg5_m = reg1_m + br;                                        \
-  reg2_m = reg0_m + bg;                                        \
-  reg3_m = reg1_m + bg;                                        \
-  reg0_m += bb;                                                \
-  reg1_m += bb;                                                \
-  vec0_m = (v8i16) __msa_ilvr_b((v16i8) zero_m, (v16i8) in1);  \
-  reg6_m = (v4i32) __msa_ilvr_h(zero_m, (v8i16) vec0_m);       \
-  reg7_m = (v4i32) __msa_ilvl_h(zero_m, (v8i16) vec0_m);       \
-  vec0_m = (v8i16) __msa_ilvr_b((v16i8) zero_m, (v16i8) in2);  \
-  reg8_m = (v4i32) __msa_ilvr_h(zero_m, (v8i16) vec0_m);       \
-  reg9_m = (v4i32) __msa_ilvl_h(zero_m, (v8i16) vec0_m);       \
-  reg0_m -= reg6_m * ub;                                       \
-  reg1_m -= reg7_m * ub;                                       \
-  reg2_m -= reg6_m * ug;                                       \
-  reg3_m -= reg7_m * ug;                                       \
-  reg4_m -= reg8_m * vr;                                       \
-  reg5_m -= reg9_m * vr;                                       \
-  reg2_m -= reg8_m * vg;                                       \
-  reg3_m -= reg9_m * vg;                                       \
-  reg0_m = __msa_srai_w(reg0_m, 6);                            \
-  reg1_m = __msa_srai_w(reg1_m, 6);                            \
-  reg2_m = __msa_srai_w(reg2_m, 6);                            \
-  reg3_m = __msa_srai_w(reg3_m, 6);                            \
-  reg4_m = __msa_srai_w(reg4_m, 6);                            \
-  reg5_m = __msa_srai_w(reg5_m, 6);                            \
-  reg0_m = __msa_maxi_s_w(reg0_m, 0);                          \
-  reg1_m = __msa_maxi_s_w(reg1_m, 0);                          \
-  reg2_m = __msa_maxi_s_w(reg2_m, 0);                          \
-  reg3_m = __msa_maxi_s_w(reg3_m, 0);                          \
-  reg4_m = __msa_maxi_s_w(reg4_m, 0);                          \
-  reg5_m = __msa_maxi_s_w(reg5_m, 0);                          \
-  reg0_m = __msa_min_s_w(reg0_m, max_val_m);                   \
-  reg1_m = __msa_min_s_w(reg1_m, max_val_m);                   \
-  reg2_m = __msa_min_s_w(reg2_m, max_val_m);                   \
-  reg3_m = __msa_min_s_w(reg3_m, max_val_m);                   \
-  reg4_m = __msa_min_s_w(reg4_m, max_val_m);                   \
-  reg5_m = __msa_min_s_w(reg5_m, max_val_m);                   \
-  out0 = __msa_pckev_h((v8i16) reg1_m, (v8i16) reg0_m);        \
-  out1 = __msa_pckev_h((v8i16) reg3_m, (v8i16) reg2_m);        \
-  out2 = __msa_pckev_h((v8i16) reg5_m, (v8i16) reg4_m);        \
+#define I422TORGB(in_y, in_u, in_v,                             \
+                  ub, vr, ug, vg, bb, bg, br, yg,               \
+                  out_b, out_g, out_r) {                        \
+  v8i16 vec0_m;                                                 \
+  v4i32 reg0_m, reg1_m, reg2_m, reg3_m, reg4_m;                 \
+  v4i32 reg5_m, reg6_m, reg7_m, reg8_m, reg9_m;                 \
+  v4i32 max_val_m = __msa_ldi_w(255);                           \
+  v8i16 zero_m = { 0 };                                         \
+                                                                \
+  in_u = (v16u8) __msa_ilvr_b((v16i8) in_u, (v16i8) in_u);      \
+  in_v = (v16u8) __msa_ilvr_b((v16i8) in_v, (v16i8) in_v);      \
+  vec0_m = (v8i16) __msa_ilvr_b((v16i8) in_y, (v16i8) in_y);    \
+  reg0_m = (v4i32) __msa_ilvr_h(zero_m, vec0_m);                \
+  reg1_m = (v4i32) __msa_ilvl_h(zero_m, vec0_m);                \
+  reg0_m *= vec_yg;                                             \
+  reg1_m *= vec_yg;                                             \
+  reg0_m = __msa_srai_w(reg0_m, 16);                            \
+  reg1_m = __msa_srai_w(reg1_m, 16);                            \
+  reg4_m = reg0_m + br;                                         \
+  reg5_m = reg1_m + br;                                         \
+  reg2_m = reg0_m + bg;                                         \
+  reg3_m = reg1_m + bg;                                         \
+  reg0_m += bb;                                                 \
+  reg1_m += bb;                                                 \
+  vec0_m = (v8i16) __msa_ilvr_b((v16i8) zero_m, (v16i8) in_u);  \
+  reg6_m = (v4i32) __msa_ilvr_h(zero_m, (v8i16) vec0_m);        \
+  reg7_m = (v4i32) __msa_ilvl_h(zero_m, (v8i16) vec0_m);        \
+  vec0_m = (v8i16) __msa_ilvr_b((v16i8) zero_m, (v16i8) in_v);  \
+  reg8_m = (v4i32) __msa_ilvr_h(zero_m, (v8i16) vec0_m);        \
+  reg9_m = (v4i32) __msa_ilvl_h(zero_m, (v8i16) vec0_m);        \
+  reg0_m -= reg6_m * ub;                                        \
+  reg1_m -= reg7_m * ub;                                        \
+  reg2_m -= reg6_m * ug;                                        \
+  reg3_m -= reg7_m * ug;                                        \
+  reg4_m -= reg8_m * vr;                                        \
+  reg5_m -= reg9_m * vr;                                        \
+  reg2_m -= reg8_m * vg;                                        \
+  reg3_m -= reg9_m * vg;                                        \
+  reg0_m = __msa_srai_w(reg0_m, 6);                             \
+  reg1_m = __msa_srai_w(reg1_m, 6);                             \
+  reg2_m = __msa_srai_w(reg2_m, 6);                             \
+  reg3_m = __msa_srai_w(reg3_m, 6);                             \
+  reg4_m = __msa_srai_w(reg4_m, 6);                             \
+  reg5_m = __msa_srai_w(reg5_m, 6);                             \
+  reg0_m = __msa_maxi_s_w(reg0_m, 0);                           \
+  reg1_m = __msa_maxi_s_w(reg1_m, 0);                           \
+  reg2_m = __msa_maxi_s_w(reg2_m, 0);                           \
+  reg3_m = __msa_maxi_s_w(reg3_m, 0);                           \
+  reg4_m = __msa_maxi_s_w(reg4_m, 0);                           \
+  reg5_m = __msa_maxi_s_w(reg5_m, 0);                           \
+  reg0_m = __msa_min_s_w(reg0_m, max_val_m);                    \
+  reg1_m = __msa_min_s_w(reg1_m, max_val_m);                    \
+  reg2_m = __msa_min_s_w(reg2_m, max_val_m);                    \
+  reg3_m = __msa_min_s_w(reg3_m, max_val_m);                    \
+  reg4_m = __msa_min_s_w(reg4_m, max_val_m);                    \
+  reg5_m = __msa_min_s_w(reg5_m, max_val_m);                    \
+  out_b = __msa_pckev_h((v8i16) reg1_m, (v8i16) reg0_m);        \
+  out_g = __msa_pckev_h((v8i16) reg3_m, (v8i16) reg2_m);        \
+  out_r = __msa_pckev_h((v8i16) reg5_m, (v8i16) reg4_m);        \
 }
 
 void MirrorRow_MSA(const uint8* src, uint8* dst, int width) {
@@ -166,13 +180,10 @@ void I422ToARGBRow_MSA(const uint8* src_y, const uint8* src_u,
                        const uint8* src_v, uint8* rgb_buf,
                        const struct YuvConstants* yuvconstants, int width) {
   int x;
-  int32 data_u, data_v;
-  int64 data_y;
   v16u8 src0, src1, src2, dst0, dst1;
   v8i16 vec0, vec1, vec2;
   v4i32 vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg, vec_br, vec_yg;
   v16u8 const_255 = (v16u8) __msa_ldi_b(255);
-  v4i32 zero = { 0 };
 
   vec_ub = __msa_fill_w(yuvconstants->kUVToB[0]);
   vec_vr = __msa_fill_w(yuvconstants->kUVToR[1]);
@@ -184,12 +195,7 @@ void I422ToARGBRow_MSA(const uint8* src_y, const uint8* src_u,
   vec_yg = __msa_fill_w(yuvconstants->kYToRgb[0]);
 
   for (x = 0; x < width; x += 8) {
-    data_y = LD(src_y);
-    data_u = LW(src_u);
-    data_v = LW(src_v);
-    src0 = (v16u8) __msa_insert_d((v2i64) zero, 0, data_y);
-    src1 = (v16u8) __msa_insert_w(zero, 0, data_u);
-    src2 = (v16u8) __msa_insert_w(zero, 0, data_v);
+    READYUV422(src_y, src_u, src_v, src0, src1, src2);
     I422TORGB(src0, src1, src2, vec_ub, vec_vr, vec_ug, vec_vg,
               vec_bb, vec_bg, vec_br, vec_yg, vec0, vec1, vec2);
     vec0 = (v8i16) __msa_ilvev_b((v16i8) vec1, (v16i8) vec0);
@@ -208,13 +214,10 @@ void I422ToRGBARow_MSA(const uint8* src_y, const uint8* src_u,
                        const uint8* src_v, uint8* rgb_buf,
                        const struct YuvConstants* yuvconstants, int width) {
   int x;
-  int64 data_y;
-  int32 data_u, data_v;
   v16u8 src0, src1, src2, dst0, dst1;
   v8i16 vec0, vec1, vec2;
   v4i32 vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg, vec_br, vec_yg;
   v16u8 const_255 = (v16u8) __msa_ldi_b(255);
-  v4i32 zero = { 0 };
 
   vec_ub = __msa_fill_w(yuvconstants->kUVToB[0]);
   vec_vr = __msa_fill_w(yuvconstants->kUVToR[1]);
@@ -226,12 +229,7 @@ void I422ToRGBARow_MSA(const uint8* src_y, const uint8* src_u,
   vec_yg = __msa_fill_w(yuvconstants->kYToRgb[0]);
 
   for (x = 0; x < width; x += 8) {
-    data_y = LD(src_y);
-    data_u = LW(src_u);
-    data_v = LW(src_v);
-    src0 = (v16u8) __msa_insert_d((v2i64) zero, 0, data_y);
-    src1 = (v16u8) __msa_insert_w(zero, 0, data_u);
-    src2 = (v16u8) __msa_insert_w(zero, 0, data_v);
+    READYUV422(src_y, src_u, src_v, src0, src1, src2);
     I422TORGB(src0, src1, src2, vec_ub, vec_vr, vec_ug, vec_vg,
               vec_bb, vec_bg, vec_br, vec_yg, vec0, vec1, vec2);
     vec0 = (v8i16) __msa_ilvev_b((v16i8) vec0, (v16i8) const_255);
@@ -246,6 +244,100 @@ void I422ToRGBARow_MSA(const uint8* src_y, const uint8* src_u,
   }
 }
 
+void I422AlphaToARGBRow_MSA(const uint8* src_y, const uint8* src_u,
+                            const uint8* src_v, const uint8* src_a,
+                            uint8* rgb_buf,
+                            const struct YuvConstants* yuvconstants,
+                            int width) {
+  int x;
+  int64 data_a;
+  v16u8 src0, src1, src2, src3, dst0, dst1;
+  v8i16 vec0, vec1, vec2;
+  v4i32 vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg, vec_br, vec_yg;
+  v4i32 zero = { 0 };
+
+  vec_ub = __msa_fill_w(yuvconstants->kUVToB[0]);
+  vec_vr = __msa_fill_w(yuvconstants->kUVToR[1]);
+  vec_ug = __msa_fill_w(yuvconstants->kUVToG[0]);
+  vec_vg = __msa_fill_w(yuvconstants->kUVToG[1]);
+  vec_bb = __msa_fill_w(yuvconstants->kUVBiasB[0]);
+  vec_bg = __msa_fill_w(yuvconstants->kUVBiasG[0]);
+  vec_br = __msa_fill_w(yuvconstants->kUVBiasR[0]);
+  vec_yg = __msa_fill_w(yuvconstants->kYToRgb[0]);
+
+  for (x = 0; x < width; x += 8) {
+    data_a = LD(src_a);
+    READYUV422(src_y, src_u, src_v, src0, src1, src2);
+    src3 = (v16u8) __msa_insert_d((v2i64) zero, 0, data_a);
+    I422TORGB(src0, src1, src2, vec_ub, vec_vr, vec_ug, vec_vg,
+              vec_bb, vec_bg, vec_br, vec_yg, vec0, vec1, vec2);
+    src3 = (v16u8) __msa_ilvr_b((v16i8) src3, (v16i8) src3);
+    vec0 = (v8i16) __msa_ilvev_b((v16i8) vec1, (v16i8) vec0);
+    vec1 = (v8i16) __msa_ilvev_b((v16i8) src3, (v16i8) vec2);
+    dst0 = (v16u8) __msa_ilvr_h((v8i16) vec1, (v8i16) vec0);
+    dst1 = (v16u8) __msa_ilvl_h((v8i16) vec1, (v8i16) vec0);
+    ST_UB2(dst0, dst1, rgb_buf, 16);
+    src_y += 8;
+    src_u += 4;
+    src_v += 4;
+    src_a += 8;
+    rgb_buf += 32;
+  }
+}
+
+void I422ToRGB24Row_MSA(const uint8* src_y, const uint8* src_u,
+                        const uint8* src_v, uint8* rgb_buf,
+                        const struct YuvConstants* yuvconstants, int32 width) {
+  int x;
+  int64 data_u, data_v;
+  v16u8 src0, src1, src2, src3, src4, src5, dst0, dst1, dst2;
+  v8i16 vec0, vec1, vec2, vec3, vec4, vec5;
+  v4i32 vec_ub, vec_vr, vec_ug, vec_vg, vec_bb, vec_bg, vec_br, vec_yg;
+  v16u8 reg0, reg1, reg2, reg3;
+  v2i64 zero = { 0 };
+  v16i8 shuffler0 = { 0, 1, 16, 2, 3, 17, 4, 5, 18, 6, 7, 19, 8, 9, 20, 10 };
+  v16i8 shuffler1 = { 0, 21, 1, 2, 22, 3, 4, 23, 5, 6, 24, 7, 8, 25, 9, 10 };
+  v16i8 shuffler2 =
+      { 26, 6, 7, 27, 8, 9, 28, 10, 11, 29, 12, 13, 30, 14, 15, 31 };
+
+  vec_ub = __msa_fill_w(yuvconstants->kUVToB[0]);
+  vec_vr = __msa_fill_w(yuvconstants->kUVToR[1]);
+  vec_ug = __msa_fill_w(yuvconstants->kUVToG[0]);
+  vec_vg = __msa_fill_w(yuvconstants->kUVToG[1]);
+  vec_bb = __msa_fill_w(yuvconstants->kUVBiasB[0]);
+  vec_bg = __msa_fill_w(yuvconstants->kUVBiasG[0]);
+  vec_br = __msa_fill_w(yuvconstants->kUVBiasR[0]);
+  vec_yg = __msa_fill_w(yuvconstants->kYToRgb[0]);
+
+  for (x = 0; x < width; x += 16) {
+    src0 = (v16u8) __msa_ld_b((v16u8*) src_y, 0);
+    data_u = LD(src_u);
+    data_v = LD(src_v);
+    src1 = (v16u8) __msa_insert_d(zero, 0, data_u);
+    src2 = (v16u8) __msa_insert_d(zero, 0, data_v);
+    src3 = (v16u8) __msa_sldi_b((v16i8) src0, (v16i8) src0, 8);
+    src4 = (v16u8) __msa_sldi_b((v16i8) src1, (v16i8) src1, 4);
+    src5 = (v16u8) __msa_sldi_b((v16i8) src2, (v16i8) src2, 4);
+    I422TORGB(src0, src1, src2, vec_ub, vec_vr, vec_ug, vec_vg,
+              vec_bb, vec_bg, vec_br, vec_yg, vec0, vec1, vec2);
+    I422TORGB(src3, src4, src5, vec_ub, vec_vr, vec_ug, vec_vg,
+              vec_bb, vec_bg, vec_br, vec_yg, vec3, vec4, vec5);
+    reg0 = (v16u8) __msa_ilvev_b((v16i8) vec1, (v16i8) vec0);
+    reg2 = (v16u8) __msa_ilvev_b((v16i8) vec4, (v16i8) vec3);
+    reg3 = (v16u8) __msa_pckev_b((v16i8) vec5, (v16i8) vec2);
+    reg1 = (v16u8) __msa_sldi_b((v16i8) reg2, (v16i8) reg0, 11);
+    dst0 = (v16u8) __msa_vshf_b(shuffler0, (v16i8) reg3, (v16i8) reg0);
+    dst1 = (v16u8) __msa_vshf_b(shuffler1, (v16i8) reg3, (v16i8) reg1);
+    dst2 = (v16u8) __msa_vshf_b(shuffler2, (v16i8) reg3, (v16i8) reg2);
+    ST_UB2(dst0, dst1, rgb_buf, 16);
+    ST_UB(dst2, (rgb_buf + 32));
+    src_y += 16;
+    src_u += 8;
+    src_v += 8;
+    rgb_buf += 48;
+  }
+}
+
 void YUY2ToYRow_MSA(const uint8* src_yuy2, uint8* dst_y, int width) {
   int x;
   v16u8 src0, src1, src2, src3, dst0, dst1;

source/row_neon64.cc

@@ -2711,7 +2711,6 @@ void SobelYRow_NEON(const uint8* src_y0, const uint8* src_y1,
   );
 }
 
-
 void HalfFloat1Row_NEON(const uint16* src, uint16* dst, float, int width) {
   asm volatile (
   "1:                                          \n"
@@ -2735,31 +2734,6 @@ void HalfFloat1Row_NEON(const uint16* src, uint16* dst, float, int width) {
   );
 }
 
-void HalfFloatRow_NEON2(const uint16* src, uint16* dst, float scale, int width) {
-  asm volatile (
-  "1:                                          \n"
-    MEMACCESS(0)
-    "ld1        {v1.16b}, [%0], #16            \n"  // load 8 shorts
-    "subs       %w2, %w2, #8                   \n"  // 8 pixels per loop
-    "uxtl       v2.4s, v1.4h                   \n"  // 8 int's
-    "uxtl2      v1.4s, v1.8h                   \n"
-    "scvtf      v2.4s, v2.4s                   \n"  // 8 floats
-    "scvtf      v1.4s, v1.4s                   \n"
-    "fmul       v2.4s, v2.4s, %3.s[0]          \n"  // adjust exponent
-    "fmul       v1.4s, v1.4s, %3.s[0]          \n"
-    "uqshrn     v4.4h, v2.4s, #13              \n"  // isolate halffloat
-    "uqshrn2    v4.8h, v1.4s, #13              \n"
-   MEMACCESS(1)
-    "st1        {v4.16b}, [%1], #16            \n"  // store 8 shorts
-    "b.gt       1b                             \n"
-  : "+r"(src),    // %0
-    "+r"(dst),    // %1
-    "+r"(width)   // %2
-  : "w"(scale * 1.9259299444e-34f)    // %3
-  : "cc", "memory", "v1", "v2", "v4"
-  );
-}
-
 void HalfFloatRow_NEON(const uint16* src, uint16* dst, float scale, int width) {
   asm volatile (
   "1:                                          \n"