diff --git a/libavfilter/x86/vf_removegrain.asm b/libavfilter/x86/vf_removegrain.asm
index 4933b72..bf2142d 100644
--- a/libavfilter/x86/vf_removegrain.asm
+++ b/libavfilter/x86/vf_removegrain.asm
@@ -68,43 +68,29 @@ SECTION_TEXT
     LOAD m8, [a8], %1
 %endmacro

-%macro SORT_AXIS 0
-    mova m9, m1
-    pminub m1, m8
-    pmaxub m8, m9
-    mova m10, m2
-    pminub m2, m7
-    pmaxub m7, m10
-    mova m11, m3
-    pminub m3, m6
-    pmaxub m6, m11
-    mova m12, m4
-    pminub m4, m5
-    pmaxub m5, m12
+; %1 data type
+; %2 simd register to hold maximums
+; %3 simd register to hold minimums
+; %4 temp location (simd register/memory)
+%macro SORT_PAIR 4
+    mova   %4, %2
+    pmin%1 %2, %3
+    pmax%1 %3, %4
 %endmacro

-%macro SORT_AXIS_16 0
-    mova m9, m1
-    pminsw m1, m8
-    pmaxsw m8, m9
-    mova m10, m2
-    pminsw m2, m7
-    pmaxsw m7, m10
-    mova m11, m3
-    pminsw m3, m6
-    pmaxsw m6, m11
-    mova m12, m4
-    pminsw m4, m5
-    pmaxsw m5, m12
+%macro SORT_AXIS 0
+    SORT_PAIR ub, m1, m8, m9
+    SORT_PAIR ub, m2, m7, m10
+    SORT_PAIR ub, m3, m6, m11
+    SORT_PAIR ub, m4, m5, m12
 %endmacro

-; %1 simd register to hold maximums
-; %2 simd register to hold minimums
-; %3 temp location (simd register/memory)
-%macro SORT_PAIR 3
-    mova %3, %1
-    pminub %1, %2
-    pmaxub %2, %3
+
+%macro SORT_AXIS_16 0
+    SORT_PAIR sw, m1, m8, m9
+    SORT_PAIR sw, m2, m7, m10
+    SORT_PAIR sw, m3, m6, m11
+    SORT_PAIR sw, m4, m5, m12
 %endmacro

 ; The loop doesn't need to do all the iterations.  It could stop when the right
@@ -115,7 +101,7 @@ SECTION_TEXT
         %assign i 1
         %assign j 2
         %rep k
-            SORT_PAIR m %+ i , m %+ j , m9
+            SORT_PAIR ub, m %+ i , m %+ j , m9
             %assign i i+1
             %assign j j+1
         %endrep
@@ -770,9 +756,9 @@ cglobal rg_fl_mode_15_16, 4, 5, 16, 0, dst, src, stride, pixels
         paddw m12, m13
         psrlw m12, 3

-        SORT_PAIR m1, m8, m0
-        SORT_PAIR m2, m7, m9
-        SORT_PAIR m3, m6, m14
+        SORT_PAIR ub, m1, m8, m0
+        SORT_PAIR ub, m2, m7, m9
+        SORT_PAIR ub, m3, m6, m14
         mova m4, m12
         mova m5, m12
         CLIPW m4, m1, m8