diff --git a/include/libyuv/row.h b/include/libyuv/row.h
index a622e3185..6519fd801 100644
--- a/include/libyuv/row.h
+++ b/include/libyuv/row.h
@@ -421,7 +421,9 @@ extern "C" {
 #define HAS_ABGRTOUVJROW_NEON
 #define HAS_ABGRTOUVROW_NEON
 #define HAS_ABGRTOYJROW_NEON
+#define HAS_ABGRTOYJROW_NEON_DOTPROD
 #define HAS_ABGRTOYROW_NEON
+#define HAS_ABGRTOYROW_NEON_DOTPROD
 #define HAS_AR64TOARGBROW_NEON
 #define HAS_ARGB1555TOARGBROW_NEON
 #define HAS_ARGB1555TOUVROW_NEON
@@ -443,12 +445,15 @@ extern "C" {
 #define HAS_ARGBTOUVJROW_NEON
 #define HAS_ARGBTOUVROW_NEON
 #define HAS_ARGBTOYJROW_NEON
+#define HAS_ARGBTOYJROW_NEON_DOTPROD
 #define HAS_ARGBTOYROW_NEON
+#define HAS_ARGBTOYROW_NEON_DOTPROD
 #define HAS_AYUVTOUVROW_NEON
 #define HAS_AYUVTOVUROW_NEON
 #define HAS_AYUVTOYROW_NEON
 #define HAS_BGRATOUVROW_NEON
 #define HAS_BGRATOYROW_NEON
+#define HAS_BGRATOYROW_NEON_DOTPROD
 #define HAS_BYTETOFLOATROW_NEON
 #define HAS_CONVERT16TO8ROW_NEON
 #define HAS_COPYROW_NEON
@@ -512,7 +517,9 @@ extern "C" {
 #define HAS_RGB565TOYROW_NEON
 #define HAS_RGBATOUVROW_NEON
 #define HAS_RGBATOYJROW_NEON
+#define HAS_RGBATOYJROW_NEON_DOTPROD
 #define HAS_RGBATOYROW_NEON
+#define HAS_RGBATOYROW_NEON_DOTPROD
 #define HAS_SETROW_NEON
 #define HAS_SPLITARGBROW_NEON
 #define HAS_SPLITRGBROW_NEON
@@ -1409,6 +1416,18 @@ void ARGBToYRow_NEON(const uint8_t* src_argb, uint8_t* dst_y, int width);
 void ARGBToYJRow_NEON(const uint8_t* src_argb, uint8_t* dst_yj, int width);
 void ABGRToYJRow_NEON(const uint8_t* src_abgr, uint8_t* dst_yj, int width);
 void RGBAToYJRow_NEON(const uint8_t* src_rgba, uint8_t* dst_yj, int width);
+void ARGBToYRow_NEON_DotProd(const uint8_t* src_argb,
+                             uint8_t* dst_y,
+                             int width);
+void ARGBToYJRow_NEON_DotProd(const uint8_t* src_argb,
+                              uint8_t* dst_yj,
+                              int width);
+void ABGRToYJRow_NEON_DotProd(const uint8_t* src_abgr,
+                              uint8_t* dst_yj,
+                              int width);
+void RGBAToYJRow_NEON_DotProd(const uint8_t* src_rgba,
+                              uint8_t* dst_yj,
+                              int width);
 void ARGBToYRow_RVV(const uint8_t* src_argb, uint8_t* dst_y, int width);
 void ARGBToYJRow_RVV(const uint8_t* src_argb, uint8_t* dst_yj, int width);
 void ABGRToYJRow_RVV(const uint8_t* src_rgba, uint8_t* dst_yj, int width);
@@ -1632,6 +1651,15 @@ void RAWToUVRow_LASX(const uint8_t* src_raw,
 void BGRAToYRow_NEON(const uint8_t* src_bgra, uint8_t* dst_y, int width);
 void ABGRToYRow_NEON(const uint8_t* src_abgr, uint8_t* dst_y, int width);
 void RGBAToYRow_NEON(const uint8_t* src_rgba, uint8_t* dst_y, int width);
+void BGRAToYRow_NEON_DotProd(const uint8_t* src_bgra,
+                             uint8_t* dst_y,
+                             int width);
+void ABGRToYRow_NEON_DotProd(const uint8_t* src_abgr,
+                             uint8_t* dst_y,
+                             int width);
+void RGBAToYRow_NEON_DotProd(const uint8_t* src_rgba,
+                             uint8_t* dst_y,
+                             int width);
 void RGB24ToYRow_NEON(const uint8_t* src_rgb24, uint8_t* dst_y, int width);
 void RGB24ToYJRow_NEON(const uint8_t* src_rgb24, uint8_t* dst_yj, int width);
 void RAWToYRow_NEON(const uint8_t* src_raw, uint8_t* dst_y, int width);
@@ -1709,12 +1737,33 @@ void RAWToYJRow_Any_SSSE3(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
 void RGB24ToYJRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
 void RAWToYJRow_Any_AVX2(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
 void ARGBToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
 void ARGBToYJRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ARGBToYJRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int width);
 void ABGRToYJRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYJRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int width);
 void RGBAToYJRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYJRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                  uint8_t* dst_ptr,
+                                  int width);
 void BGRAToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void BGRAToYRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
 void ABGRToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void ABGRToYRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
 void RGBAToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
+void RGBAToYRow_Any_NEON_DotProd(const uint8_t* src_ptr,
+                                 uint8_t* dst_ptr,
+                                 int width);
 void RGB24ToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
 void RGB24ToYJRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
 void RAWToYRow_Any_NEON(const uint8_t* src_ptr, uint8_t* dst_ptr, int width);
diff --git a/source/convert.cc b/source/convert.cc
index 2aa1865ac..e852a90cd 100644
--- a/source/convert.cc
+++ b/source/convert.cc
@@ -1904,6 +1904,14 @@ int ARGBToI420(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ARGBToUVRow = ARGBToUVRow_Any_NEON;
@@ -2066,6 +2074,14 @@ int ARGBToI420Alpha(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ARGBToUVRow = ARGBToUVRow_Any_NEON;
@@ -2229,6 +2245,14 @@ int BGRAToI420(const uint8_t* src_bgra,
     }
   }
 #endif
+#if defined(HAS_BGRATOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    BGRAToYRow = BGRAToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      BGRAToYRow = BGRAToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_BGRATOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     BGRAToUVRow = BGRAToUVRow_Any_NEON;
@@ -2388,6 +2412,14 @@ int ABGRToI420(const uint8_t* src_abgr,
     }
   }
 #endif
+#if defined(HAS_ABGRTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYRow = ABGRToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ABGRTOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ABGRToUVRow = ABGRToUVRow_Any_NEON;
@@ -2497,6 +2529,14 @@ int RGBAToI420(const uint8_t* src_rgba,
     }
   }
 #endif
+#if defined(HAS_RGBATOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    RGBAToYRow = RGBAToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYRow = RGBAToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_RGBATOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     RGBAToUVRow = RGBAToUVRow_Any_NEON;
diff --git a/source/convert_from_argb.cc b/source/convert_from_argb.cc
index b45de8c8f..6c361c05a 100644
--- a/source/convert_from_argb.cc
+++ b/source/convert_from_argb.cc
@@ -116,6 +116,14 @@ int ARGBToI444(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOYROW_MSA)
   if (TestCpuFlag(kCpuHasMSA)) {
     ARGBToYRow = ARGBToYRow_Any_MSA;
@@ -231,6 +239,14 @@ int ARGBToI422(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ARGBToUVRow = ARGBToUVRow_Any_NEON;
@@ -331,6 +347,14 @@ int ARGBToNV12(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ARGBToUVRow = ARGBToUVRow_Any_NEON;
@@ -553,6 +577,14 @@ int ARGBToNV21(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ARGBToUVRow = ARGBToUVRow_Any_NEON;
@@ -752,6 +784,14 @@ int ABGRToNV12(const uint8_t* src_abgr,
     }
   }
 #endif
+#if defined(HAS_ABGRTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYRow = ABGRToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ABGRTOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ABGRToUVRow = ABGRToUVRow_Any_NEON;
@@ -940,6 +980,14 @@ int ABGRToNV21(const uint8_t* src_abgr,
     }
   }
 #endif
+#if defined(HAS_ABGRTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYRow = ABGRToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYRow = ABGRToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ABGRTOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ABGRToUVRow = ABGRToUVRow_Any_NEON;
@@ -1133,6 +1181,14 @@ int ARGBToYUY2(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ARGBToUVRow = ARGBToUVRow_Any_NEON;
@@ -1329,6 +1385,14 @@ int ARGBToUYVY(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOUVROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ARGBToUVRow = ARGBToUVRow_Any_NEON;
@@ -1501,6 +1565,14 @@ int ARGBToI400(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYRow = ARGBToYRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYRow = ARGBToYRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOYROW_MSA)
   if (TestCpuFlag(kCpuHasMSA)) {
     ARGBToYRow = ARGBToYRow_Any_MSA;
@@ -2248,6 +2320,14 @@ int ARGBToJ420(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYJRow = ARGBToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOUVJROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ARGBToUVJRow = ARGBToUVJRow_Any_NEON;
@@ -2416,6 +2496,14 @@ int ARGBToJ422(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYJRow = ARGBToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOUVJROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ARGBToUVJRow = ARGBToUVJRow_Any_NEON;
@@ -2522,6 +2610,14 @@ int ARGBToJ400(const uint8_t* src_argb,
     }
   }
 #endif
+#if defined(HAS_ARGBTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ARGBToYJRow = ARGBToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ARGBToYJRow = ARGBToYJRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ARGBTOYJROW_MSA)
   if (TestCpuFlag(kCpuHasMSA)) {
     ARGBToYJRow = ARGBToYJRow_Any_MSA;
@@ -2593,6 +2689,14 @@ int RGBAToJ400(const uint8_t* src_rgba,
     }
   }
 #endif
+#if defined(HAS_RGBATOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    RGBAToYJRow = RGBAToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      RGBAToYJRow = RGBAToYJRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_RGBATOYJROW_MSA)
   if (TestCpuFlag(kCpuHasMSA)) {
     RGBAToYJRow = RGBAToYJRow_Any_MSA;
@@ -2698,6 +2802,14 @@ int ABGRToJ420(const uint8_t* src_abgr,
     }
   }
 #endif
+#if defined(HAS_ABGRTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYJRow = ABGRToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ABGRTOUVJROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ABGRToUVJRow = ABGRToUVJRow_Any_NEON;
@@ -2828,6 +2940,14 @@ int ABGRToJ422(const uint8_t* src_abgr,
     }
   }
 #endif
+#if defined(HAS_ABGRTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYJRow = ABGRToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ABGRTOUVJROW_NEON)
   if (TestCpuFlag(kCpuHasNEON)) {
     ABGRToUVJRow = ABGRToUVJRow_Any_NEON;
@@ -2930,6 +3050,14 @@ int ABGRToJ400(const uint8_t* src_abgr,
     }
   }
 #endif
+#if defined(HAS_ABGRTOYJROW_NEON_DOTPROD)
+  if (TestCpuFlag(kCpuHasNeonDotProd)) {
+    ABGRToYJRow = ABGRToYJRow_Any_NEON_DotProd;
+    if (IS_ALIGNED(width, 16)) {
+      ABGRToYJRow = ABGRToYJRow_NEON_DotProd;
+    }
+  }
+#endif
 #if defined(HAS_ABGRTOYJROW_MSA)
   if (TestCpuFlag(kCpuHasMSA)) {
     ABGRToYJRow = ABGRToYJRow_Any_MSA;
diff --git a/source/row_any.cc b/source/row_any.cc
index e574543cc..8ed5a49c2 100644
--- a/source/row_any.cc
+++ b/source/row_any.cc
@@ -1072,6 +1072,9 @@ ANY11(RGBAToYJRow_Any_SSSE3, RGBAToYJRow_SSSE3, 0, 4, 1, 15)
 #ifdef HAS_ARGBTOYROW_NEON
 ANY11(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 0, 4, 1, 15)
 #endif
+#ifdef HAS_ARGBTOYROW_NEON_DOTPROD
+ANY11(ARGBToYRow_Any_NEON_DotProd, ARGBToYRow_NEON_DotProd, 0, 4, 1, 15)
+#endif
 #ifdef HAS_ARGBTOYROW_MSA
 ANY11(ARGBToYRow_Any_MSA, ARGBToYRow_MSA, 0, 4, 1, 15)
 #endif
@@ -1084,12 +1087,21 @@ ANY11(ARGBToYRow_Any_LASX, ARGBToYRow_LASX, 0, 4, 1, 31)
 #ifdef HAS_ARGBTOYJROW_NEON
 ANY11(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, 0, 4, 1, 15)
 #endif
+#ifdef HAS_ARGBTOYJROW_NEON_DOTPROD
+ANY11(ARGBToYJRow_Any_NEON_DotProd, ARGBToYJRow_NEON_DotProd, 0, 4, 1, 15)
+#endif
 #ifdef HAS_ABGRTOYJROW_NEON
 ANY11(ABGRToYJRow_Any_NEON, ABGRToYJRow_NEON, 0, 4, 1, 15)
 #endif
+#ifdef HAS_ABGRTOYJROW_NEON_DOTPROD
+ANY11(ABGRToYJRow_Any_NEON_DotProd, ABGRToYJRow_NEON_DotProd, 0, 4, 1, 15)
+#endif
 #ifdef HAS_RGBATOYJROW_NEON
 ANY11(RGBAToYJRow_Any_NEON, RGBAToYJRow_NEON, 0, 4, 1, 15)
 #endif
+#ifdef HAS_RGBATOYJROW_NEON_DOTPROD
+ANY11(RGBAToYJRow_Any_NEON_DotProd, RGBAToYJRow_NEON_DotProd, 0, 4, 1, 15)
+#endif
 #ifdef HAS_ARGBTOYJROW_MSA
 ANY11(ARGBToYJRow_Any_MSA, ARGBToYJRow_MSA, 0, 4, 1, 15)
 #endif
@@ -1114,6 +1126,9 @@ ANY11(ABGRToYJRow_Any_LASX, ABGRToYJRow_LASX, 0, 4, 1, 31)
 #ifdef HAS_BGRATOYROW_NEON
 ANY11(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 0, 4, 1, 15)
 #endif
+#ifdef HAS_BGRATOYROW_NEON_DOTPROD
+ANY11(BGRAToYRow_Any_NEON_DotProd, BGRAToYRow_NEON_DotProd, 0, 4, 1, 15)
+#endif
 #ifdef HAS_BGRATOYROW_MSA
 ANY11(BGRAToYRow_Any_MSA, BGRAToYRow_MSA, 0, 4, 1, 15)
 #endif
@@ -1126,6 +1141,9 @@ ANY11(BGRAToYRow_Any_LASX, BGRAToYRow_LASX, 0, 4, 1, 31)
 #ifdef HAS_ABGRTOYROW_NEON
 ANY11(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 0, 4, 1, 15)
 #endif
+#ifdef HAS_ABGRTOYROW_NEON_DOTPROD
+ANY11(ABGRToYRow_Any_NEON_DotProd, ABGRToYRow_NEON_DotProd, 0, 4, 1, 15)
+#endif
 #ifdef HAS_ABGRTOYROW_MSA
 ANY11(ABGRToYRow_Any_MSA, ABGRToYRow_MSA, 0, 4, 1, 7)
 #endif
@@ -1138,6 +1156,9 @@ ANY11(ABGRToYRow_Any_LASX, ABGRToYRow_LASX, 0, 4, 1, 31)
 #ifdef HAS_RGBATOYROW_NEON
 ANY11(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 0, 4, 1, 15)
 #endif
+#ifdef HAS_RGBATOYROW_NEON_DOTPROD
+ANY11(RGBAToYRow_Any_NEON_DotProd, RGBAToYRow_NEON_DotProd, 0, 4, 1, 15)
+#endif
 #ifdef HAS_RGBATOYROW_MSA
 ANY11(RGBAToYRow_Any_MSA, RGBAToYRow_MSA, 0, 4, 1, 15)
 #endif
diff --git a/source/row_neon64.cc b/source/row_neon64.cc
index 4920292cf..e06f65e56 100644
--- a/source/row_neon64.cc
+++ b/source/row_neon64.cc
@@ -3029,12 +3029,49 @@ void ARGBToYMatrixRow_NEON(const uint8_t* src_argb,
         "v17");
 }
 
+void
+ARGBToYMatrixRow_NEON_DotProd(const uint8_t* src_argb,
+                              uint8_t* dst_y,
+                              int width,
+                              const struct RgbConstants* rgbconstants) {
+  asm volatile(
+      "ldr         d0, [%3]                      \n"  // load rgbconstants
+      "dup         v16.4s, v0.s[0]               \n"
+      "dup         v17.8h,  v0.h[2]              \n"
+      "1:                                        \n"
+      "ld1         {v4.16b, v5.16b, v6.16b, v7.16b}, [%0], #64 \n"  // load 16
+                                                                    // pixels.
+      "subs        %w2, %w2, #16                 \n"  // 16 processed per loop.
+      "movi        v0.16b, #0                    \n"
+      "movi        v1.16b, #0                    \n"
+      "movi        v2.16b, #0                    \n"
+      "movi        v3.16b, #0                    \n"
+      "udot        v0.4s, v4.16b, v16.16b        \n"
+      "udot        v1.4s, v5.16b, v16.16b        \n"
+      "udot        v2.4s, v6.16b, v16.16b        \n"
+      "udot        v3.4s, v7.16b, v16.16b        \n"
+      "uzp1        v0.8h, v0.8h, v1.8h           \n"
+      "uzp1        v1.8h, v2.8h, v3.8h           \n"
+      "addhn       v0.8b, v0.8h, v17.8h          \n"
+      "addhn       v1.8b, v1.8h, v17.8h          \n"
+      "st1         {v0.8b, v1.8b}, [%1], #16     \n"  // store 16 pixels Y.
+      "b.gt        1b                            \n"
+      : "+r"(src_argb),    // %0
+        "+r"(dst_y),       // %1
+        "+r"(width)        // %2
+      : "r"(rgbconstants)  // %3
+      : "cc", "memory", "v0", "v1", "v2", "v3", "v4", "v5", "v6", "v7", "v16",
+        "v17");
+}
+
 // RGB to JPeg coefficients
 // B * 0.1140 coefficient = 29
 // G * 0.5870 coefficient = 150
 // R * 0.2990 coefficient = 77
 // Add 0.5 = 0x80
 static const struct RgbConstants kRgb24JPEGConstants = {{29, 150, 77, 0}, 128};
+static const struct RgbConstants kRgb24JPEGDotProdConstants = {{0, 29, 150, 77},
+                                                               128};
 
 static const struct RgbConstants kRawJPEGConstants = {{77, 150, 29, 0}, 128};
 
@@ -3046,8 +3083,12 @@ static const struct RgbConstants kRawJPEGConstants = {{77, 150, 29, 0}, 128};
 
 static const struct RgbConstants kRgb24I601Constants = {{25, 129, 66, 0},
                                                         0x1080};
+static const struct RgbConstants kRgb24I601DotProdConstants = {{0, 25, 129, 66},
+                                                               0x1080};
 
 static const struct RgbConstants kRawI601Constants = {{66, 129, 25, 0}, 0x1080};
+static const struct RgbConstants kRawI601DotProdConstants = {{0, 66, 129, 25},
+                                                             0x1080};
 
 void ARGBToYRow_NEON(const uint8_t* src_argb, uint8_t* dst_y, int width) {
   ARGBToYMatrixRow_NEON(src_argb, dst_y, width, &kRgb24I601Constants);
@@ -3065,6 +3106,30 @@ void ABGRToYJRow_NEON(const uint8_t* src_abgr, uint8_t* dst_yj, int width) {
   ARGBToYMatrixRow_NEON(src_abgr, dst_yj, width, &kRawJPEGConstants);
 }
 
+void ARGBToYRow_NEON_DotProd(const uint8_t* src_argb,
+                             uint8_t* dst_y,
+                             int width) {
+  ARGBToYMatrixRow_NEON_DotProd(src_argb, dst_y, width, &kRgb24I601Constants);
+}
+
+void ARGBToYJRow_NEON_DotProd(const uint8_t* src_argb,
+                              uint8_t* dst_yj,
+                              int width) {
+  ARGBToYMatrixRow_NEON_DotProd(src_argb, dst_yj, width, &kRgb24JPEGConstants);
+}
+
+void ABGRToYRow_NEON_DotProd(const uint8_t* src_abgr,
+                             uint8_t* dst_y,
+                             int width) {
+  ARGBToYMatrixRow_NEON_DotProd(src_abgr, dst_y, width, &kRawI601Constants);
+}
+
+void ABGRToYJRow_NEON_DotProd(const uint8_t* src_abgr,
+                              uint8_t* dst_yj,
+                              int width) {
+  ARGBToYMatrixRow_NEON_DotProd(src_abgr, dst_yj, width, &kRawJPEGConstants);
+}
+
 // RGBA expects first value to be A and ignored, then 3 values to contain RGB.
 // Same code as ARGB, except the LD4
 void RGBAToYMatrixRow_NEON(const uint8_t* src_rgba,
@@ -3112,6 +3177,33 @@ void BGRAToYRow_NEON(const uint8_t* src_bgra, uint8_t* dst_y, int width) {
   RGBAToYMatrixRow_NEON(src_bgra, dst_y, width, &kRawI601Constants);
 }
 
+void RGBAToYRow_NEON_DotProd(const uint8_t* src_rgba,
+                             uint8_t* dst_y,
+                             int width) {
+  // No need for a separate implementation for RGBA inputs, just permute the
+  // RGB constants.
+  ARGBToYMatrixRow_NEON_DotProd(src_rgba, dst_y, width,
+                                &kRgb24I601DotProdConstants);
+}
+
+void RGBAToYJRow_NEON_DotProd(const uint8_t* src_rgba,
+                              uint8_t* dst_yj,
+                              int width) {
+  // No need for a separate implementation for RGBA inputs, just permute the
+  // RGB constants.
+  ARGBToYMatrixRow_NEON_DotProd(src_rgba, dst_yj, width,
+                                &kRgb24JPEGDotProdConstants);
+}
+
+void BGRAToYRow_NEON_DotProd(const uint8_t* src_bgra,
+                             uint8_t* dst_y,
+                             int width) {
+  // No need for a separate implementation for RGBA inputs, just permute the
+  // RGB constants.
+  ARGBToYMatrixRow_NEON_DotProd(src_bgra, dst_y, width,
+                                &kRawI601DotProdConstants);
+}
+
 void RGBToYMatrixRow_NEON(const uint8_t* src_rgb,
                           uint8_t* dst_y,
                           int width,