pub/ed25519.c: Rearrange `ptadd' to use fewer registers.

[catacomb] / pub / ed25519.c

diff --git a/pub/ed25519.c b/pub/ed25519.c
index 830e5de..c6b805c 100644 (file)
--- a/pub/ed25519.c
+++ b/pub/ed25519.c
@@ -172,7 +172,7 @@ static void ptadd(f25519 *X, f25519 *Y, f25519 *Z,
                  const f25519 *X0, const f25519 *Y0, const f25519 *Z0,
                  const f25519 *X1, const f25519 *Y1, const f25519 *Z1)
 {
-  f25519 t0, t1, t2, t3, t4, t5;
+  f25519 t0, t1, t2, t3;
 
   /* Bernstein, Birkner, Joye, Lange, and Peters, `Twisted Edwards Curves',
    * 2008-03-13, https://cr.yp.to/newelliptic/twisted-20080313.pdf shows the
@@ -187,23 +187,23 @@ static void ptadd(f25519 *X, f25519 *Y, f25519 *Z,
    */
 
   f25519_mul(&t0, Z0, Z1);             /* t0 = A = Z0 Z1 */
-  f25519_sqr(&t1, &t0);                        /* t1 = B = A^2 */
+  f25519_add(&t1, X0, Y0);             /* t1 = X0 + Y0 */
+  f25519_add(&t2, X1, Y1);             /* t2 = X1 + Y1 */
+  f25519_mul(&t1, &t1, &t2);           /* t1 = (X0 + Y0) (X1 + Y1) */
   f25519_mul(&t2, X0, X1);             /* t2 = C = X0 X1 */
   f25519_mul(&t3, Y0, Y1);             /* t3 = D = Y0 Y1 */
-  f25519_mul(&t4, &t2, &t3);           /* t4 = C D */
-  f25519_mul(&t4, &t4, D);             /* t4 = E = d C D */
-  f25519_sub(&t5, &t1, &t4);           /* t5 = F = B - E */
-  f25519_add(&t4, &t1, &t4);           /* t4 = G = B + E */
-  f25519_add(&t1, &t2, &t3);           /* t1 = C + D */
-  f25519_add(&t2, X0, Y0);             /* t2 = X0 + Y0 */
-  f25519_add(&t3, X1, Y1);             /* t3 = X1 + Y1 */
-  f25519_mul(X, &t0, &t5);             /* X = A F */
-  f25519_mul(Y, &t0, &t4);             /* Y = A G */
-  f25519_mul(Z, &t5, &t4);             /* Z = F G */
-  f25519_mul(Y, Y, &t1);               /* Y = A G (C + D) = A G (D - a C) */
-  f25519_mul(&t0, &t2, &t3);           /* t0 = (X0 + Y0) (X1 + Y1) */
-  f25519_sub(&t0, &t0, &t1);          /* t0 = (X0 + Y0) (X1 + Y1) - C - D */
-  f25519_mul(X, X, &t0);         /* X = A F ((X0 + Y0) (X1 + Y1) - C - D) */
+  f25519_add(Y, &t2, &t3);             /* Y = C + D = D - a C */
+  f25519_sub(X, &t1, Y);               /* X = (X0 + Y0) (X1 + Y1) - C - D */
+  f25519_mul(X, X, &t0);           /* X = A ((X0 + Y0) (X1 + Y1) - C - D) */
+  f25519_mul(Y, Y, &t0);               /* Y = A (D - a C) */
+  f25519_sqr(&t0, &t0);                        /* t0 = B = A^2 */
+  f25519_mul(&t1, &t2, &t3);           /* t1 = C D */
+  f25519_mul(&t1, &t1, D);             /* t1 = E = d C D */
+  f25519_sub(&t2, &t0, &t1);           /* t2 = F = B - E */
+  f25519_add(&t1, &t0, &t1);           /* t1 = G = B + E */
+  f25519_mul(X, X, &t2);         /* X = A F ((X0 + Y0) (X1 + Y1) - C - D) */
+  f25519_mul(Y, Y, &t1);               /* Y = A G (D - a C) */
+  f25519_mul(Z, &t1, &t2);             /* Z = F G */
 }
 
 static void ptdbl(f25519 *X, f25519 *Y, f25519 *Z,
@@ -389,7 +389,7 @@ static void unpack_key(scaf_piece a[NPIECE], octet h1[32],
   sha512_init(&h); sha512_hash(&h, k, ksz); sha512_done(&h, b);
   b[0] &= 0xf8u; b[31] = (b[31]&0x3f) | 0x40;
   scaf_load(a, b, 32, NPIECE, PIECEWD);
-  memcpy(h1, b + 32, 32);
+  if (h1) memcpy(h1, b + 32, 32);
 }
 
 /*----- Main code ---------------------------------------------------------*/
@@ -409,9 +409,8 @@ void ed25519_pubkey(octet K[ED25519_PUBSZ], const void *k, size_t ksz)
 {
   scaf_piece a[NPIECE];
   f25519 AX, AY, AZ;
-  octet h1[32];
 
-  unpack_key(a, h1, k, ksz);
+  unpack_key(a, 0, k, ksz);
   ptmul(&AX, &AY, &AZ, a, BX, BY, BZ);
   ptencode(K, &AX, &AY, &AZ);
 }