diff --git a/thiersort.h b/thiersort.h
index f7207ff..1e9b827 100644
--- a/thiersort.h
+++ b/thiersort.h
@@ -426,23 +426,25 @@ static inline void thiersort_apply(
 		/* Replace the whole chain starting at i (j for chain index) */
 		/* This works because we can easily check what does'nt need moves anymore. */
 		TSU32 j = i;
-		TSU32 ni = sortres[j].i;
+		TSU32 ni = sortres[j].i; /* (*) */
 
 		/* Until would move to its own position, chain */
 		/* This also solves "already at right place" originals. */
 		while(ni != j) {
-			/* xchg j and ni */
-			memcpy(tmp, &((TSU8*)arr)[(size_t)j * elemsize], elemsize);
-			memcpy(
-				&((TSU8*)arr)[(size_t)j * elemsize],
-				&((TSU8*)arr)[(size_t)ni * elemsize],
-				elemsize);
-			memcpy(&((TSU8*)arr)[(size_t)ni * elemsize], tmp, elemsize);
+			/* xchg j and ni - but only if not already swapped! See: (*) */
+			if(sortres[ni].i != ni) {
+				memcpy(tmp, &((TSU8*)arr)[(size_t)j * elemsize], elemsize);
+				memcpy(
+					&((TSU8*)arr)[(size_t)j * elemsize],
+					&((TSU8*)arr)[(size_t)ni * elemsize],
+					elemsize);
+				memcpy(&((TSU8*)arr)[(size_t)ni * elemsize], tmp, elemsize);
+			}
 
 			/* Mark j index as done in sortres for outer loop. */
 			/* This is necessary for inner loop stopping early */
 			/* and outer catch up on already processed location. */
-			sortres[j].i = j;
+			sortres[j].i = j; /* (*) */
 
 			/* Update j and ni */
 			/* Now we must find what should be at new location j */
@@ -451,6 +453,9 @@ static inline void thiersort_apply(
 			ni = sortres[j].i;
 		}
 	}
+
+	/* Release mem - better we do because we changed it behind the scenes! */
+	free(sortres);
 }
 
 /**