diff --git a/third_party/lcms/src/cmsopt.c b/third_party/lcms/src/cmsopt.c
index 5ea1b4c85..ea93cd413 100644
--- a/third_party/lcms/src/cmsopt.c
+++ b/third_party/lcms/src/cmsopt.c
@@ -104,6 +104,18 @@ typedef struct {
 // Simple optimizations ----------------------------------------------------------------------------------------------------------
 
 
+// Perform one row of matrix multiply with translation for MatShaperEval16().
+cmsINLINE cmsInt64Number _MatShaperEvaluateRow(cmsS1Fixed14Number* mat,
+                                               cmsS1Fixed14Number off,
+                                               cmsS1Fixed14Number r,
+                                               cmsS1Fixed14Number g,
+                                               cmsS1Fixed14Number b) {
+  return ((cmsInt64Number)mat[0] * r +
+          (cmsInt64Number)mat[1] * g +
+          (cmsInt64Number)mat[2] * b +
+          off + 0x2000) >> 14;
+}
+
 // Remove an element in linked chain
 static
 void _RemoveElement(cmsStage** head)
@@ -1527,7 +1539,8 @@ void MatShaperEval16(register const cmsUInt16Number In[],
                      register const void* D)
 {
     MatShaper8Data* p = (MatShaper8Data*) D;
-    cmsS1Fixed14Number l1, l2, l3, r, g, b;
+    cmsS1Fixed14Number r, g, b;
+    cmsInt64Number l1, l2, l3;
     cmsUInt32Number ri, gi, bi;
 
     // In this case (and only in this case!) we can use this simplification since
@@ -1542,9 +1555,9 @@ void MatShaperEval16(register const cmsUInt16Number In[],
     b = p->Shaper1B[bi];
 
     // Evaluate the matrix in 1.14 fixed point
-    l1 =  (p->Mat[0][0] * r + p->Mat[0][1] * g + p->Mat[0][2] * b + p->Off[0] + 0x2000) >> 14;
-    l2 =  (p->Mat[1][0] * r + p->Mat[1][1] * g + p->Mat[1][2] * b + p->Off[1] + 0x2000) >> 14;
-    l3 =  (p->Mat[2][0] * r + p->Mat[2][1] * g + p->Mat[2][2] * b + p->Off[2] + 0x2000) >> 14;
+    l1 = _MatShaperEvaluateRow(p->Mat[0], p->Off[0], r, g, b);
+    l2 = _MatShaperEvaluateRow(p->Mat[1], p->Off[1], r, g, b);
+    l3 = _MatShaperEvaluateRow(p->Mat[2], p->Off[2], r, g, b);
 
     // Now we have to clip to 0..1.0 range
     ri = (l1 < 0) ? 0 : ((l1 > 16384) ? 16384U : (cmsUInt32Number) l1);