Horribly mangle scrypt.c

amd64 linux speedup from 2.02 to 2.67 kH/s with default options, from 2.59 to 3.24kH/s with -O3

1 files changed, 155 insertions, 283 deletions

diff --git a/scrypt.c b/scrypt.c
index 3b80a81..ebe3477 100644
--- a/scrypt.c
+++ b/scrypt.c
@@ -55,26 +55,6 @@ be32enc(void *pp, uint32_t x)
 	p[0] = (x >> 24) & 0xff;
 }
 
-static inline uint32_t
-le32dec(const void *pp)
-{
-	const uint8_t *p = (uint8_t const *)pp;
-
-	return ((uint32_t)(p[0]) + ((uint32_t)(p[1]) << 8) +
-	    ((uint32_t)(p[2]) << 16) + ((uint32_t)(p[3]) << 24));
-}
-
-static inline void
-le32enc(void *pp, uint32_t x)
-{
-	uint8_t * p = (uint8_t *)pp;
-
-	p[0] = x & 0xff;
-	p[1] = (x >> 8) & 0xff;
-	p[2] = (x >> 16) & 0xff;
-	p[3] = (x >> 24) & 0xff;
-}
-
 
 typedef struct SHA256Context {
 	uint32_t state[8];
@@ -91,7 +71,7 @@ typedef struct HMAC_SHA256Context {
  * Encode a length len/4 vector of (uint32_t) into a length len vector of
  * (unsigned char) in big-endian form.  Assumes len is a multiple of 4.
  */
-static void
+static inline void
 be32enc_vect(unsigned char *dst, const uint32_t *src, size_t len)
 {
 	size_t i;
@@ -104,7 +84,7 @@ be32enc_vect(unsigned char *dst, const uint32_t *src, size_t len)
  * Decode a big-endian length len vector of (unsigned char) into a length
  * len/4 vector of (uint32_t).  Assumes len is a multiple of 4.
  */
-static void
+static inline void
 be32dec_vect(uint32_t *dst, const unsigned char *src, size_t len)
 {
 	size_t i;
@@ -227,11 +207,6 @@ SHA256_Transform(uint32_t * state, const unsigned char block[64])
 	/* 4. Mix local working variables into global state */
 	for (i = 0; i < 8; i++)
 		state[i] += S[i];
-
-	/* Clean the stack. */
-	memset(W, 0, 256);
-	memset(S, 0, 32);
-	t0 = t1 = 0;
 }
 
 static unsigned char PAD[64] = {
@@ -242,7 +217,7 @@ static unsigned char PAD[64] = {
 };
 
 /* SHA-256 initialization.  Begins a SHA-256 operation. */
-static void
+static inline void
 SHA256_Init(SHA256_CTX * ctx)
 {
 
@@ -261,7 +236,7 @@ SHA256_Init(SHA256_CTX * ctx)
 }
 
 /* Add bytes into the hash */
-static void
+static inline void
 SHA256_Update(SHA256_CTX * ctx, const void *in, size_t len)
 {
 	uint32_t bitlen[2];
@@ -304,7 +279,7 @@ SHA256_Update(SHA256_CTX * ctx, const void *in, size_t len)
 }
 
 /* Add padding and terminating bit-count. */
-static void
+static inline void
 SHA256_Pad(SHA256_CTX * ctx)
 {
 	unsigned char len[8];
@@ -329,7 +304,7 @@ SHA256_Pad(SHA256_CTX * ctx)
  * SHA-256 finalization.  Pads the input data, exports the hash value,
  * and clears the context state.
  */
-static void
+static inline void
 SHA256_Final(unsigned char digest[32], SHA256_CTX * ctx)
 {
 
@@ -338,141 +313,108 @@ SHA256_Final(unsigned char digest[32], SHA256_CTX * ctx)
 
 	/* Write the hash */
 	be32enc_vect(digest, ctx->state, 32);
-
-	/* Clear the context state */
-	memset((void *)ctx, 0, sizeof(*ctx));
 }
 
-/* Initialize an HMAC-SHA256 operation with the given key. */
-static void
-HMAC_SHA256_Init(HMAC_SHA256_CTX * ctx, const void * _K, size_t Klen)
+/**
+ * PBKDF2_SHA256(passwd, passwdlen, salt, saltlen, c, buf, dkLen):
+ * Compute PBKDF2(passwd, salt, c, dkLen) using HMAC-SHA256 as the PRF, and
+ * write the output to buf.  The value dkLen must be at most 32 * (2^32 - 1).
+ */
+static inline void
+PBKDF2_SHA256_80_128(const uint8_t * passwd, uint8_t * buf)
 {
+	HMAC_SHA256_CTX PShctx, hctx;
+	size_t i;
+	uint8_t ivec[4];
+	unsigned char ihash[32];
+
+	/* Compute HMAC state after processing P and S. */
 	unsigned char pad[64];
 	unsigned char khash[32];
-	const unsigned char * K = _K;
-	size_t i;
 
 	/* If Klen > 64, the key is really SHA256(K). */
-	if (Klen > 64) {
-		SHA256_Init(&ctx->ictx);
-		SHA256_Update(&ctx->ictx, K, Klen);
-		SHA256_Final(khash, &ctx->ictx);
-		K = khash;
-		Klen = 32;
-	}
+	SHA256_Init(&PShctx.ictx);
+	SHA256_Update(&PShctx.ictx, passwd, 80);
+	SHA256_Final(khash, &PShctx.ictx);
 
-	/* Inner SHA256 operation is SHA256(K xor [block of 0x36] || data). */
-	SHA256_Init(&ctx->ictx);
+	SHA256_Init(&PShctx.ictx);
 	memset(pad, 0x36, 64);
-	for (i = 0; i < Klen; i++)
-		pad[i] ^= K[i];
-	SHA256_Update(&ctx->ictx, pad, 64);
+	for (i = 0; i < 32; i++)
+		pad[i] ^= khash[i];
+	SHA256_Update(&PShctx.ictx, pad, 64);
 
-	/* Outer SHA256 operation is SHA256(K xor [block of 0x5c] || hash). */
-	SHA256_Init(&ctx->octx);
+	SHA256_Init(&PShctx.octx);
 	memset(pad, 0x5c, 64);
-	for (i = 0; i < Klen; i++)
-		pad[i] ^= K[i];
-	SHA256_Update(&ctx->octx, pad, 64);
+	for (i = 0; i < 32; i++)
+		pad[i] ^= khash[i];
+	SHA256_Update(&PShctx.octx, pad, 64);
 
-	/* Clean the stack. */
-	memset(khash, 0, 32);
-}
+	SHA256_Update(&PShctx.ictx, passwd, 80);
 
-/* Add bytes to the HMAC-SHA256 operation. */
-static void
-HMAC_SHA256_Update(HMAC_SHA256_CTX * ctx, const void *in, size_t len)
-{
-
-	/* Feed data to the inner SHA256 operation. */
-	SHA256_Update(&ctx->ictx, in, len);
-}
-
-/* Finish an HMAC-SHA256 operation. */
-static void
-HMAC_SHA256_Final(unsigned char digest[32], HMAC_SHA256_CTX * ctx)
-{
-	unsigned char ihash[32];
-
-	/* Finish the inner SHA256 operation. */
-	SHA256_Final(ihash, &ctx->ictx);
-
-	/* Feed the inner hash to the outer SHA256 operation. */
-	SHA256_Update(&ctx->octx, ihash, 32);
+	/* Iterate through the blocks. */
+	for (i = 0; i * 32 < 128; i++) {
+		/* Generate INT(i + 1). */
+		be32enc(ivec, (uint32_t)(i + 1));
 
-	/* Finish the outer SHA256 operation. */
-	SHA256_Final(digest, &ctx->octx);
+		/* Compute U_1 = PRF(P, S || INT(i)). */
+		memcpy(&hctx, &PShctx, sizeof(HMAC_SHA256_CTX));
+		SHA256_Update(&hctx.ictx, ivec, 4);
 
-	/* Clean the stack. */
-	memset(ihash, 0, 32);
+		SHA256_Final(ihash, &hctx.ictx);
+		/* Feed the inner hash to the outer SHA256 operation. */
+		SHA256_Update(&hctx.octx, ihash, 32);
+		/* Finish the outer SHA256 operation. */
+		SHA256_Final(&buf[i*32], &hctx.octx);
+	}
 }
 
-/**
- * PBKDF2_SHA256(passwd, passwdlen, salt, saltlen, c, buf, dkLen):
- * Compute PBKDF2(passwd, salt, c, dkLen) using HMAC-SHA256 as the PRF, and
- * write the output to buf.  The value dkLen must be at most 32 * (2^32 - 1).
- */
-static void
-PBKDF2_SHA256(const uint8_t * passwd, size_t passwdlen, const uint8_t * salt,
-    size_t saltlen, uint64_t c, uint8_t * buf, size_t dkLen)
+
+static inline void
+PBKDF2_SHA256_80_128_32(const uint8_t * passwd, const uint8_t * salt, uint8_t * buf)
 {
-	HMAC_SHA256_CTX PShctx, hctx;
+	HMAC_SHA256_CTX PShctx;
 	size_t i;
 	uint8_t ivec[4];
-	uint8_t U[32];
-	uint8_t T[32];
-	uint64_t j;
-	int k;
-	size_t clen;
+	unsigned char ihash[32];
 
 	/* Compute HMAC state after processing P and S. */
-	HMAC_SHA256_Init(&PShctx, passwd, passwdlen);
-	HMAC_SHA256_Update(&PShctx, salt, saltlen);
+	unsigned char pad[64];
+	unsigned char khash[32];
 
-	/* Iterate through the blocks. */
-	for (i = 0; i * 32 < dkLen; i++) {
-		/* Generate INT(i + 1). */
-		be32enc(ivec, (uint32_t)(i + 1));
+	/* If Klen > 64, the key is really SHA256(K). */
+	SHA256_Init(&PShctx.ictx);
+	SHA256_Update(&PShctx.ictx, passwd, 80);
+	SHA256_Final(khash, &PShctx.ictx);
 
-		/* Compute U_1 = PRF(P, S || INT(i)). */
-		memcpy(&hctx, &PShctx, sizeof(HMAC_SHA256_CTX));
-		HMAC_SHA256_Update(&hctx, ivec, 4);
-		HMAC_SHA256_Final(U, &hctx);
+	SHA256_Init(&PShctx.ictx);
+	memset(pad, 0x36, 64);
+	for (i = 0; i < 32; i++)
+		pad[i] ^= khash[i];
+	SHA256_Update(&PShctx.ictx, pad, 64);
 
-		/* T_i = U_1 ... */
-		memcpy(T, U, 32);
+	SHA256_Init(&PShctx.octx);
+	memset(pad, 0x5c, 64);
+	for (i = 0; i < 32; i++)
+		pad[i] ^= khash[i];
+	SHA256_Update(&PShctx.octx, pad, 64);
 
-		for (j = 2; j <= c; j++) {
-			/* Compute U_j. */
-			HMAC_SHA256_Init(&hctx, passwd, passwdlen);
-			HMAC_SHA256_Update(&hctx, U, 32);
-			HMAC_SHA256_Final(U, &hctx);
+	SHA256_Update(&PShctx.ictx, salt, 128);
 
-			/* ... xor U_j ... */
-			for (k = 0; k < 32; k++)
-				T[k] ^= U[k];
-		}
+	/* Generate INT(i + 1). */
+	be32enc(ivec, (uint32_t)(1));
 
-		/* Copy as many bytes as necessary into buf. */
-		clen = dkLen - i * 32;
-		if (clen > 32)
-			clen = 32;
-		memcpy(&buf[i * 32], T, clen);
-	}
+	/* Compute U_1 = PRF(P, S || INT(i)). */
+	SHA256_Update(&PShctx.ictx, ivec, 4);
 
-	/* Clean PShctx, since we never called _Final on it. */
-	memset(&PShctx, 0, sizeof(HMAC_SHA256_CTX));
+	SHA256_Final(ihash, &PShctx.ictx);
+	/* Feed the inner hash to the outer SHA256 operation. */
+	SHA256_Update(&PShctx.octx, ihash, 32);
+	/* Finish the outer SHA256 operation. */
+	SHA256_Final(&buf[0], &PShctx.octx);
 }
 
 
-static void blkcpy(void *, void *, size_t);
-static void blkxor(void *, void *, size_t);
-static void salsa20_8(uint32_t[16]);
-static void blockmix_salsa8(uint32_t *, uint32_t *, uint32_t *, size_t);
-static uint64_t integerify(void *, size_t);
-static void smix(uint8_t *, size_t, uint64_t, uint32_t *, uint32_t *);
-
-static void
+static inline void
 blkcpy(void * dest, void * src, size_t len)
 {
 	size_t * D = dest;
@@ -484,7 +426,7 @@ blkcpy(void * dest, void * src, size_t len)
 		D[i] = S[i];
 }
 
-static void
+static inline void
 blkxor(void * dest, void * src, size_t len)
 {
 	size_t * D = dest;
@@ -500,149 +442,59 @@ blkxor(void * dest, void * src, size_t len)
  * salsa20_8(B):
  * Apply the salsa20/8 core to the provided block.
  */
-static void
+static inline void
 salsa20_8(uint32_t B[16])
 {
-	uint32_t x[16];
+	uint32_t x00,x01,x02,x03,x04,x05,x06,x07,x08,x09,x10,x11,x12,x13,x14,x15;
 	size_t i;
 
-	blkcpy(x, B, 64);
+	x00 = B[ 0];
+	x01 = B[ 1];
+	x02 = B[ 2];
+	x03 = B[ 3];
+	x04 = B[ 4];
+	x05 = B[ 5];
+	x06 = B[ 6];
+	x07 = B[ 7];
+	x08 = B[ 8];
+	x09 = B[ 9];
+	x10 = B[10];
+	x11 = B[11];
+	x12 = B[12];
+	x13 = B[13];
+	x14 = B[14];
+	x15 = B[15];
 	for (i = 0; i < 8; i += 2) {
 #define R(a,b) (((a) << (b)) | ((a) >> (32 - (b))))
 		/* Operate on columns. */
-		x[ 4] ^= R(x[ 0]+x[12], 7);  x[ 8] ^= R(x[ 4]+x[ 0], 9);
-		x[12] ^= R(x[ 8]+x[ 4],13);  x[ 0] ^= R(x[12]+x[ 8],18);
-
-		x[ 9] ^= R(x[ 5]+x[ 1], 7);  x[13] ^= R(x[ 9]+x[ 5], 9);
-		x[ 1] ^= R(x[13]+x[ 9],13);  x[ 5] ^= R(x[ 1]+x[13],18);
-
-		x[14] ^= R(x[10]+x[ 6], 7);  x[ 2] ^= R(x[14]+x[10], 9);
-		x[ 6] ^= R(x[ 2]+x[14],13);  x[10] ^= R(x[ 6]+x[ 2],18);
-
-		x[ 3] ^= R(x[15]+x[11], 7);  x[ 7] ^= R(x[ 3]+x[15], 9);
-		x[11] ^= R(x[ 7]+x[ 3],13);  x[15] ^= R(x[11]+x[ 7],18);
+		x04 ^= R(x00+x12, 7);	x08 ^= R(x04+x00, 9);	x12 ^= R(x08+x04,13);	x00 ^= R(x12+x08,18);
+		x09 ^= R(x05+x01, 7);	x13 ^= R(x09+x05, 9);	x01 ^= R(x13+x09,13);	x05 ^= R(x01+x13,18);
+		x14 ^= R(x10+x06, 7);	x02 ^= R(x14+x10, 9);	x06 ^= R(x02+x14,13);	x10 ^= R(x06+x02,18);
+		x03 ^= R(x15+x11, 7);	x07 ^= R(x03+x15, 9);	x11 ^= R(x07+x03,13);	x15 ^= R(x11+x07,18);
 
 		/* Operate on rows. */
-		x[ 1] ^= R(x[ 0]+x[ 3], 7);  x[ 2] ^= R(x[ 1]+x[ 0], 9);
-		x[ 3] ^= R(x[ 2]+x[ 1],13);  x[ 0] ^= R(x[ 3]+x[ 2],18);
-
-		x[ 6] ^= R(x[ 5]+x[ 4], 7);  x[ 7] ^= R(x[ 6]+x[ 5], 9);
-		x[ 4] ^= R(x[ 7]+x[ 6],13);  x[ 5] ^= R(x[ 4]+x[ 7],18);
-
-		x[11] ^= R(x[10]+x[ 9], 7);  x[ 8] ^= R(x[11]+x[10], 9);
-		x[ 9] ^= R(x[ 8]+x[11],13);  x[10] ^= R(x[ 9]+x[ 8],18);
-
-		x[12] ^= R(x[15]+x[14], 7);  x[13] ^= R(x[12]+x[15], 9);
-		x[14] ^= R(x[13]+x[12],13);  x[15] ^= R(x[14]+x[13],18);
+		x01 ^= R(x00+x03, 7);	x02 ^= R(x01+x00, 9);	x03 ^= R(x02+x01,13);	x00 ^= R(x03+x02,18);
+		x06 ^= R(x05+x04, 7);	x07 ^= R(x06+x05, 9);	x04 ^= R(x07+x06,13);	x05 ^= R(x04+x07,18);
+		x11 ^= R(x10+x09, 7);	x08 ^= R(x11+x10, 9);	x09 ^= R(x08+x11,13);	x10 ^= R(x09+x08,18);
+		x12 ^= R(x15+x14, 7);	x13 ^= R(x12+x15, 9);	x14 ^= R(x13+x12,13);	x15 ^= R(x14+x13,18);
 #undef R
 	}
-	for (i = 0; i < 16; i++)
-		B[i] += x[i];
-}
-
-/**
- * blockmix_salsa8(Bin, Bout, X, r):
- * Compute Bout = BlockMix_{salsa20/8, r}(Bin).  The input Bin must be 128r
- * bytes in length; the output Bout must also be the same size.  The
- * temporary space X must be 64 bytes.
- */
-static void
-blockmix_salsa8(uint32_t * Bin, uint32_t * Bout, uint32_t * X, size_t r)
-{
-	size_t i;
-
-	/* 1: X <-- B_{2r - 1} */
-	blkcpy(X, &Bin[(2 * r - 1) * 16], 64);
-
-	/* 2: for i = 0 to 2r - 1 do */
-	for (i = 0; i < 2 * r; i += 2) {
-		/* 3: X <-- H(X \xor B_i) */
-		blkxor(X, &Bin[i * 16], 64);
-		salsa20_8(X);
-
-		/* 4: Y_i <-- X */
-		/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
-		blkcpy(&Bout[i * 8], X, 64);
-
-		/* 3: X <-- H(X \xor B_i) */
-		blkxor(X, &Bin[i * 16 + 16], 64);
-		salsa20_8(X);
-
-		/* 4: Y_i <-- X */
-		/* 6: B' <-- (Y_0, Y_2 ... Y_{2r-2}, Y_1, Y_3 ... Y_{2r-1}) */
-		blkcpy(&Bout[i * 8 + r * 16], X, 64);
-	}
-}
-
-/**
- * integerify(B, r):
- * Return the result of parsing B_{2r-1} as a little-endian integer.
- */
-static uint64_t
-integerify(void * B, size_t r)
-{
-	uint32_t * X = (void *)((uintptr_t)(B) + (2 * r - 1) * 64);
-
-	return (((uint64_t)(X[1]) << 32) + X[0]);
-}
-
-/**
- * smix(B, r, N, V, XY):
- * Compute B = SMix_r(B, N).  The input B must be 128r bytes in length;
- * the temporary storage V must be 128rN bytes in length; the temporary
- * storage XY must be 256r + 64 bytes in length.  The value N must be a
- * power of 2 greater than 1.  The arrays B, V, and XY must be aligned to a
- * multiple of 64 bytes.
- */
-static void
-smix(uint8_t * B, size_t r, uint64_t N, uint32_t * V, uint32_t * XY)
-{
-	uint32_t * X = XY;
-	uint32_t * Y = &XY[32 * r];
-	uint32_t * Z = &XY[64 * r];
-	uint64_t i;
-	uint64_t j;
-	size_t k;
-
-	/* 1: X <-- B */
-	for (k = 0; k < 32 * r; k++)
-		X[k] = le32dec(&B[4 * k]);
-
-	/* 2: for i = 0 to N - 1 do */
-	for (i = 0; i < N; i += 2) {
-		/* 3: V_i <-- X */
-		blkcpy(&V[i * (32 * r)], X, 128 * r);
-
-		/* 4: X <-- H(X) */
-		blockmix_salsa8(X, Y, Z, r);
-
-		/* 3: V_i <-- X */
-		blkcpy(&V[(i + 1) * (32 * r)], Y, 128 * r);
-
-		/* 4: X <-- H(X) */
-		blockmix_salsa8(Y, X, Z, r);
-	}
-
-	/* 6: for i = 0 to N - 1 do */
-	for (i = 0; i < N; i += 2) {
-		/* 7: j <-- Integerify(X) mod N */
-		j = integerify(X, r) & (N - 1);
-
-		/* 8: X <-- H(X \xor V_j) */
-		blkxor(X, &V[j * (32 * r)], 128 * r);
-		blockmix_salsa8(X, Y, Z, r);
-
-		/* 7: j <-- Integerify(X) mod N */
-		j = integerify(Y, r) & (N - 1);
-
-		/* 8: X <-- H(X \xor V_j) */
-		blkxor(Y, &V[j * (32 * r)], 128 * r);
-		blockmix_salsa8(Y, X, Z, r);
-	}
-
-	/* 10: B' <-- X */
-	for (k = 0; k < 32 * r; k++)
-		le32enc(&B[4 * k], X[k]);
+	B[ 0] += x00;
+	B[ 1] += x01;
+	B[ 2] += x02;
+	B[ 3] += x03;
+	B[ 4] += x04;
+	B[ 5] += x05;
+	B[ 6] += x06;
+	B[ 7] += x07;
+	B[ 8] += x08;
+	B[ 9] += x09;
+	B[10] += x10;
+	B[11] += x11;
+	B[12] += x12;
+	B[13] += x13;
+	B[14] += x14;
+	B[15] += x15;
 }
 
 /* cpu and memory intensive function to transform a 80 byte buffer into a 32 byte output
@@ -650,30 +502,50 @@ smix(uint8_t * B, size_t r, uint64_t N, uint32_t * V, uint32_t * XY)
  */
 static void scrypt_1024_1_1_256_sp(const char* input, char* output, char* scratchpad)
 {
-	uint8_t * B;
 	uint32_t * V;
-	uint32_t * XY;
+	uint32_t * X;
 	uint32_t i;
+	uint32_t j;
+
+	X = (uint32_t *)(((uintptr_t)(scratchpad) + 63) & ~ (uintptr_t)(63));
+	V = &X[32];
 
-	const uint32_t N = 1024;
-	const uint32_t r = 1;
-	const uint32_t p = 1;
+	PBKDF2_SHA256_80_128((const uint8_t*)input, (uint8_t *)X);
 
-	B = (uint8_t *)(((uintptr_t)(scratchpad) + 63) & ~ (uintptr_t)(63));
-	XY = (uint32_t *)(B + (128 * r * p));
-	V = (uint32_t *)(B + (128 * r * p) + (256 * r + 64));
+	for (i = 0; i < 1024; i += 2) {
+		blkcpy(&V[i * 32], X, 128);
 
-	/* 1: (B_0 ... B_{p-1}) <-- PBKDF2(P, S, 1, p * MFLen) */
-	PBKDF2_SHA256((const uint8_t*)input, 80, (const uint8_t*)input, 80, 1, B, p * 128 * r);
+		blkxor(&X[0], &X[16], 64);
+		salsa20_8(&X[0]);
+		blkxor(&X[16], &X[0], 64);
+		salsa20_8(&X[16]);
 
-	/* 2: for i = 0 to p - 1 do */
-	for (i = 0; i < p; i++) {
-		/* 3: B_i <-- MF(B_i, N) */
-		smix(&B[i * 128 * r], r, N, V, XY);
+		blkcpy(&V[(i + 1) * 32], X, 128);
+
+		blkxor(&X[0], &X[16], 64);
+		salsa20_8(&X[0]);
+		blkxor(&X[16], &X[0], 64);
+		salsa20_8(&X[16]);
+	}
+	for (i = 0; i < 1024; i += 2) {
+		j = X[16] & 1023;
+		blkxor(X, &V[j * 32], 128);
+
+		blkxor(&X[0], &X[16], 64);
+		salsa20_8(&X[0]);
+		blkxor(&X[16], &X[0], 64);
+		salsa20_8(&X[16]);
+
+		j = X[16] & 1023;
+		blkxor(X, &V[j * 32], 128);
+
+		blkxor(&X[0], &X[16], 64);
+		salsa20_8(&X[0]);
+		blkxor(&X[16], &X[0], 64);
+		salsa20_8(&X[16]);
 	}
 
-	/* 5: DK <-- PBKDF2(P, B, 1, dkLen) */
-	PBKDF2_SHA256((const uint8_t*)input, 80, B, p * 128 * r, 1, (uint8_t*)output, 32);
+	PBKDF2_SHA256_80_128_32((const uint8_t*)input, (const uint8_t *)X, (uint8_t*)output);
 }
 
 int scanhash_scrypt(int thr_id, unsigned char *pdata, unsigned char *scratchbuf,