diff --git a/ed25519.dll b/ed25519_32.dll
similarity index 87%
rename from ed25519.dll
rename to ed25519_32.dll
index 33f9938..d5077a1 100644
Binary files a/ed25519.dll and b/ed25519_32.dll differ
diff --git a/ed25519_64.dll b/ed25519_64.dll
new file mode 100644
index 0000000..78c8199
Binary files /dev/null and b/ed25519_64.dll differ
diff --git a/src/add_scalar.c b/src/add_scalar.c
index 91b64d3..d728995 100644
--- a/src/add_scalar.c
+++ b/src/add_scalar.c
@@ -10,7 +10,7 @@ void ed25519_add_scalar(unsigned char *public_key, unsigned char *private_key, c
     unsigned char n[32]; 
     ge_p3 nB;
     ge_p1p1 A_p1p1;
-    ge_p2 A;
+    ge_p3 A;
     ge_p3 public_key_unpacked;
     ge_cached T;
 
@@ -27,21 +27,30 @@ void ed25519_add_scalar(unsigned char *public_key, unsigned char *private_key, c
         sc_muladd(private_key, SC_1, n, private_key);
     }
 
+	/* public key: A = nB + T */
     if (public_key) {
-        /* unpack public key into T */
-        ge_frombytes_negate_vartime(&public_key_unpacked, public_key);
-        fe_neg(public_key_unpacked.X, public_key_unpacked.X); // undo negate
-        fe_neg(public_key_unpacked.T, public_key_unpacked.T); // undo negate
-        ge_p3_to_cached(&T, &public_key_unpacked);
+		/* if we know the private key we don't need a point addition, which is faster */
+		/* using a "timing attack" you could find out wether or not we know the private
+		   key, but this information seems rather useless - if this is important pass
+		   public_key and private_key seperately in 2 function calls */
+		if (private_key) {
+			ge_scalarmult_base(&A, private_key);
+		} else {
+			/* unpack public key into T */
+			ge_frombytes_negate_vartime(&public_key_unpacked, public_key);
+			fe_neg(public_key_unpacked.X, public_key_unpacked.X); // undo negate
+			fe_neg(public_key_unpacked.T, public_key_unpacked.T); // undo negate
+			ge_p3_to_cached(&T, &public_key_unpacked);
 
-        /* calculate n*B */
-        ge_scalarmult_base(&nB, n);
+			/* calculate n*B */
+			ge_scalarmult_base(&nB, n);
 
-        /* A = n*B + T */
-        ge_add(&A_p1p1, &nB, &T);
-        ge_p1p1_to_p2(&A, &A_p1p1);
-        
-        /* pack public key */
-        ge_tobytes(public_key, &A);
+			/* A = n*B + T */
+			ge_add(&A_p1p1, &nB, &T);
+			ge_p1p1_to_p3(&A, &A_p1p1);
+		}
+			
+		/* pack public key */
+		ge_p3_tobytes(public_key, &A);
     }
 }
diff --git a/test.c b/test.c
index 264832f..b317438 100644
--- a/test.c
+++ b/test.c
@@ -80,15 +80,24 @@ int main(int argc, char *argv[]) {
 	}
 
     /* test performance */
+	printf("testing seed generation performance: ");
+    start = clock();
+    for (i = 0; i < 10000; ++i) {
+		ed25519_create_seed(seed);
+    }
+    end = clock();
+
+    printf("%fus per seed\n", ((double) ((end - start) * 1000)) / CLOCKS_PER_SEC / i * 1000);
+
+
     printf("testing key generation performance: ");
     start = clock();
     for (i = 0; i < 10000; ++i) {
-        ed25519_create_seed(seed);
         ed25519_create_keypair(public_key, private_key, seed);
     }
     end = clock();
 
-    printf("%fus per seed and keypair\n", ((double) ((end - start) * 1000)) / CLOCKS_PER_SEC / i * 1000);
+    printf("%fus per keypair\n", ((double) ((end - start) * 1000)) / CLOCKS_PER_SEC / i * 1000);
 
     printf("testing sign performance: ");
     start = clock();
@@ -109,7 +118,7 @@ int main(int argc, char *argv[]) {
     printf("%fus per signature\n", ((double) ((end - start) * 1000)) / CLOCKS_PER_SEC / i * 1000);
     
 
-    printf("testing scalar addition performance: ");
+    printf("testing keypair scalar addition performance: ");
     start = clock();
     for (i = 0; i < 10000; ++i) {
         ed25519_add_scalar(public_key, private_key, scalar);
@@ -118,6 +127,15 @@ int main(int argc, char *argv[]) {
 
     printf("%fus per keypair\n", ((double) ((end - start) * 1000)) / CLOCKS_PER_SEC / i * 1000);
 
+	printf("testing public key scalar addition performance: ");
+    start = clock();
+    for (i = 0; i < 10000; ++i) {
+        ed25519_add_scalar(public_key, NULL, scalar);
+    }
+    end = clock();
+
+    printf("%fus per key\n", ((double) ((end - start) * 1000)) / CLOCKS_PER_SEC / i * 1000);
+
 	printf("testing key exchange performance: ");
     start = clock();
     for (i = 0; i < 10000; ++i) {