diff --git a/source/scale_common.cc b/source/scale_common.cc
index 3507aa4d9..e9c4eff93 100644
--- a/source/scale_common.cc
+++ b/source/scale_common.cc
@@ -421,7 +421,7 @@ void ScaleColsUp2_16_C(uint16* dst_ptr, const uint16* src_ptr,
 #define BLENDER(a, b, f) (uint8)((int)(a) + \
     ((((int)((f)) * ((int)(b) - (int)(a))) + 0x8000) >> 16))
 #else
-// inteluses 7 bit math with rounding.
+// Intel uses 7 bit math with rounding.
 #define BLENDER(a, b, f) (uint8)((int)(a) + \
     (((int)((f) >> 9) * ((int)(b) - (int)(a)) + 0x40) >> 7))
 #endif
diff --git a/source/scale_gcc.cc b/source/scale_gcc.cc
index e2f88544b..1ab5a7627 100644
--- a/source/scale_gcc.cc
+++ b/source/scale_gcc.cc
@@ -867,7 +867,7 @@ void ScaleFilterCols_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
     "pshufb    %%xmm5,%%xmm1                   \n"
     "punpcklwd %%xmm4,%%xmm0                   \n"
     "psubb     %8,%%xmm0                       \n"  // make pixels signed.
-    "pxor      %%xmm6,%%xmm1                   \n"  // 128 -f = (f ^ 127 ) + 1
+    "pxor      %%xmm6,%%xmm1                   \n"  // 128 - f = (f ^ 127 ) + 1
     "paddusb   %%xmm7,%%xmm1                   \n"
     "pmaddubsw %%xmm0,%%xmm1                   \n"
     "pextrw    $0x1,%%xmm2,%k3                 \n"
diff --git a/source/scale_neon.cc b/source/scale_neon.cc
index 44b0c8080..cb0a6ee5c 100644
--- a/source/scale_neon.cc
+++ b/source/scale_neon.cc
@@ -572,9 +572,9 @@ void ScaleAddRows_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
     MEMACCESS(6)                                               \
     "vld2.8     {d6["#n"], d7["#n"]}, [%6]     \n"
 
-// The NEON version mimics this formula:
-// #define BLENDER(a, b, f) (uint8)((int)(a) +
-//    ((int)(f) * ((int)(b) - (int)(a)) >> 16))
+// The NEON version mimics this formula (from row_common.cc):
+// #define BLENDER(a, b, f) (uint8)((int)(a) + \
+//    ((((int)((f)) * ((int)(b) - (int)(a))) + 0x8000) >> 16))
 
 void ScaleFilterCols_NEON(uint8* dst_ptr, const uint8* src_ptr,
                           int dst_width, int x, int dx) {
diff --git a/source/scale_neon64.cc b/source/scale_neon64.cc
index ff277f26f..2362f0874 100644
--- a/source/scale_neon64.cc
+++ b/source/scale_neon64.cc
@@ -587,6 +587,10 @@ void ScaleAddRows_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
     MEMACCESS(6)                                               \
     "ld2        {v4.b, v5.b}["#n"], [%6]      \n"
 
+// The NEON version mimics this formula (from row_common.cc):
+// #define BLENDER(a, b, f) (uint8)((int)(a) + \
+//    ((((int)((f)) * ((int)(b) - (int)(a))) + 0x8000) >> 16))
+
 void ScaleFilterCols_NEON(uint8* dst_ptr, const uint8* src_ptr,
                           int dst_width, int x, int dx) {
   int dx_offset[4] = {0, 1, 2, 3};
@@ -626,8 +630,8 @@ void ScaleFilterCols_NEON(uint8* dst_ptr, const uint8* src_ptr,
     "ushll2    v6.4s, v6.8h, #0                \n"
     "mul       v16.4s, v16.4s, v7.4s           \n"
     "mul       v17.4s, v17.4s, v6.4s           \n"
-    "rshrn      v6.4h, v16.4s, #16             \n"
-    "rshrn2     v6.8h, v17.4s, #16             \n"
+    "rshrn     v6.4h, v16.4s, #16              \n"
+    "rshrn2    v6.8h, v17.4s, #16              \n"
     "add       v4.8h, v4.8h, v6.8h             \n"
     "xtn       v4.8b, v4.8h                    \n"