Cleanup, slightly slow down V1 identity generation, faster Salsa20 (by a bit).

node/Identity.cpp

@@ -127,10 +127,10 @@ bool identityV1ProofOfWorkCriteria(const void *in, const unsigned int len)
 	std::sort(w, w + (ZT_IDENTITY_V1_POW_MEMORY_SIZE / 8), p_CompareLittleEndian());
 	Poly1305::compute(w, w, ZT_IDENTITY_V1_POW_MEMORY_SIZE, w);
 
-	// PoW criteria passed if this is true. The value 593 was chosen experimentally
+	// PoW criteria passed if this is true. The value 1093 was chosen experimentally
 	// to yield a good average performance balancing fast setup with intentional
 	// identity collision resistance.
-	return (Utils::ntoh(w[0]) % 593U) == 0;
+	return (Utils::ntoh(w[0]) % 1000U) == 0;
 }
 
 } // anonymous namespace

node/Salsa20.cpp

@@ -10,66 +10,54 @@
 #include "Constants.hpp"
 #include "Salsa20.hpp"
 
-#define ROTATE(v,c) (((v) << (c)) | ((v) >> (32 - (c))))
-#define XOR(v,w) ((v) ^ (w))
-#define PLUS(v,w) ((uint32_t)((v) + (w)))
+#define ROTATE(v, c) (((v) << (c)) | ((v) >> (32 - (c))))
+#define XOR(v, w) ((v) ^ (w))
+#define PLUS(v, w) ((uint32_t)((v) + (w)))
 
-// Set up load/store macros with appropriate endianness (we don't use these in SSE mode)
 #ifndef ZT_SALSA20_SSE
-
 #if __BYTE_ORDER == __LITTLE_ENDIAN
-
 #ifdef ZT_NO_UNALIGNED_ACCESS
 // Slower version that does not use type punning
 #define U8TO32_LITTLE(p) ( ((uint32_t)(p)[0]) | ((uint32_t)(p)[1] << 8) | ((uint32_t)(p)[2] << 16) | ((uint32_t)(p)[3] << 24) )
-static inline void U32TO8_LITTLE(uint8_t *const c,const uint32_t v) { c[0] = (uint8_t)v; c[1] = (uint8_t)(v >> 8); c[2] = (uint8_t)(v >> 16); c[3] = (uint8_t)(v >> 24); }
+static ZT_INLINE void U32TO8_LITTLE(uint8_t *const c,const uint32_t v) { c[0] = (uint8_t)v; c[1] = (uint8_t)(v >> 8); c[2] = (uint8_t)(v >> 16); c[3] = (uint8_t)(v >> 24); }
 #else
 // Fast version that just does 32-bit load/store
 #define U8TO32_LITTLE(p) (*((const uint32_t *)((const void *)(p))))
 #define U32TO8_LITTLE(c,v) *((uint32_t *)((void *)(c))) = (v)
 #endif // ZT_NO_UNALIGNED_ACCESS
-
 #else // __BYTE_ORDER == __BIG_ENDIAN (we don't support anything else... does MIDDLE_ENDIAN even still exist?)
-
 #ifdef __GNUC__
-
 // Use GNUC builtin bswap macros on big-endian machines if available
 #define U8TO32_LITTLE(p) __builtin_bswap32(*((const uint32_t *)((const void *)(p))))
 #define U32TO8_LITTLE(c,v) *((uint32_t *)((void *)(c))) = __builtin_bswap32((v))
-
 #else // no __GNUC__
-
 // Otherwise do it the slow, manual way on BE machines
 #define U8TO32_LITTLE(p) ( ((uint32_t)(p)[0]) | ((uint32_t)(p)[1] << 8) | ((uint32_t)(p)[2] << 16) | ((uint32_t)(p)[3] << 24) )
-static inline void U32TO8_LITTLE(uint8_t *const c,const uint32_t v) { c[0] = (uint8_t)v; c[1] = (uint8_t)(v >> 8); c[2] = (uint8_t)(v >> 16); c[3] = (uint8_t)(v >> 24); }
-
+static ZT_INLINE void U32TO8_LITTLE(uint8_t *const c,const uint32_t v) { c[0] = (uint8_t)v; c[1] = (uint8_t)(v >> 8); c[2] = (uint8_t)(v >> 16); c[3] = (uint8_t)(v >> 24); }
 #endif // __GNUC__ or not
-
 #endif // __BYTE_ORDER little or big?
-
 #endif // !ZT_SALSA20_SSE
 
-// Statically compute and define SSE constants
 #ifdef ZT_SALSA20_SSE
 class _s20sseconsts
 {
 public:
-	_s20sseconsts()
+	_s20sseconsts() noexcept
 	{
 		maskLo32 = _mm_shuffle_epi32(_mm_cvtsi32_si128(-1), _MM_SHUFFLE(1, 0, 1, 0));
 		maskHi32 = _mm_slli_epi64(maskLo32, 32);
 	}
-	__m128i maskLo32,maskHi32;
+	__m128i maskLo32, maskHi32;
 };
-static const _s20sseconsts _S20SSECONSTANTS;
+static const _s20sseconsts s_S20SSECONSTANTS;
 #endif
 
 namespace ZeroTier {
 
-void Salsa20::init(const void *key,const void *iv) noexcept
+void Salsa20::init(const void *key, const void *iv) noexcept
 {
 #ifdef ZT_SALSA20_SSE
-	const uint32_t *const k = (const uint32_t *)key;
+	const uint32_t *const k = (const uint32_t *) key;
 	_state.i[0] = 0x61707865;
 	_state.i[1] = 0x3320646e;
 	_state.i[2] = 0x79622d32;
@@ -81,10 +69,10 @@ void Salsa20::init(const void *key,const void *iv) noexcept
 	_state.i[8] = 0;
 	_state.i[9] = k[6];
 	_state.i[10] = k[1];
-	_state.i[11] = ((const uint32_t *)iv)[1];
+	_state.i[11] = ((const uint32_t *) iv)[1];
 	_state.i[12] = k[5];
 	_state.i[13] = k[0];
-	_state.i[14] = ((const uint32_t *)iv)[0];
+	_state.i[14] = ((const uint32_t *) iv)[0];
 	_state.i[15] = k[4];
 #else
 	const char *const constants = "expand 32-byte k";
@@ -108,44 +96,58 @@ void Salsa20::init(const void *key,const void *iv) noexcept
 #endif
 }
 
-void Salsa20::crypt12(const void *in,void *out,unsigned int bytes) noexcept
+union p_SalsaState {
+#ifdef ZT_SALSA20_SSE
+	__m128i v[4];
+#endif // ZT_SALSA20_SSE
+	uint32_t i[16];
+};
+
+template<unsigned int R>
+static ZT_INLINE void p_salsaCrypt(p_SalsaState *const state, const uint8_t *m, uint8_t *c, unsigned int bytes) noexcept
 {
+	if (unlikely(bytes == 0))
+		return;
+
 	uint8_t tmp[64];
-	const uint8_t *m = (const uint8_t *)in;
-	uint8_t *c = (uint8_t *)out;
 	uint8_t *ctarget = c;
-	unsigned int i;
 
-#ifndef ZT_SALSA20_SSE
+#ifdef ZT_SALSA20_SSE
+	_mm_prefetch(m, _MM_HINT_T0);
+	_mm_prefetch(m + 64, _MM_HINT_T0);
+	__m128i X0 = state->v[0];
+	__m128i X1 = state->v[1];
+	__m128i X2 = state->v[2];
+	__m128i X3 = state->v[3];
+	const __m128i maskLo32 = s_S20SSECONSTANTS.maskLo32;
+	const __m128i maskHi32 = s_S20SSECONSTANTS.maskHi32;
+	const __m128i add1 = _mm_set_epi32(0, 0, 0, 1);
+#else
 	uint32_t x0, x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15;
 	uint32_t j0, j1, j2, j3, j4, j5, j6, j7, j8, j9, j10, j11, j12, j13, j14, j15;
-#endif
-
-	if (!bytes)
-		return;
-
-#ifndef ZT_SALSA20_SSE
-	j0 = _state.i[0];
-	j1 = _state.i[1];
-	j2 = _state.i[2];
-	j3 = _state.i[3];
-	j4 = _state.i[4];
-	j5 = _state.i[5];
-	j6 = _state.i[6];
-	j7 = _state.i[7];
-	j8 = _state.i[8];
-	j9 = _state.i[9];
-	j10 = _state.i[10];
-	j11 = _state.i[11];
-	j12 = _state.i[12];
-	j13 = _state.i[13];
-	j14 = _state.i[14];
-	j15 = _state.i[15];
+	j0 = state->i[0];
+	j1 = state->i[1];
+	j2 = state->i[2];
+	j3 = state->i[3];
+	j4 = state->i[4];
+	j5 = state->i[5];
+	j6 = state->i[6];
+	j7 = state->i[7];
+	j8 = state->i[8];
+	j9 = state->i[9];
+	j10 = state->i[10];
+	j11 = state->i[11];
+	j12 = state->i[12];
+	j13 = state->i[13];
+	j14 = state->i[14];
+	j15 = state->i[15];
 #endif
 
 	for (;;) {
-		if (bytes < 64) {
-			for (i = 0;i < bytes;++i)
+		if (likely(bytes >= 64)) {
+			_mm_prefetch(m + 128, _MM_HINT_T0);
+		} else {
+			for (unsigned int i = 0;i < bytes;++i)
 				tmp[i] = m[i];
 			m = tmp;
 			ctarget = c;
@@ -153,179 +155,61 @@ void Salsa20::crypt12(const void *in,void *out,unsigned int bytes) noexcept
 		}
 
 #ifdef ZT_SALSA20_SSE
-		__m128i X0 = _mm_loadu_si128((const __m128i *)&(_state.v[0]));
-		__m128i X1 = _mm_loadu_si128((const __m128i *)&(_state.v[1]));
-		__m128i X2 = _mm_loadu_si128((const __m128i *)&(_state.v[2]));
-		__m128i X3 = _mm_loadu_si128((const __m128i *)&(_state.v[3]));
-		__m128i T;
 		__m128i X0s = X0;
 		__m128i X1s = X1;
 		__m128i X2s = X2;
 		__m128i X3s = X3;
+		__m128i T;
 
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
+		for(unsigned int rr=0; rr<(R/2); ++rr) {
+			T = _mm_add_epi32(X0, X3);
+			X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
+			T = _mm_add_epi32(X1, X0);
+			X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
+			T = _mm_add_epi32(X2, X1);
+			X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
+			T = _mm_add_epi32(X3, X2);
+			X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
+			X1 = _mm_shuffle_epi32(X1, 0x93);
+			X2 = _mm_shuffle_epi32(X2, 0x4E);
+			X3 = _mm_shuffle_epi32(X3, 0x39);
+			T = _mm_add_epi32(X0, X1);
+			X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
+			T = _mm_add_epi32(X3, X0);
+			X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
+			T = _mm_add_epi32(X2, X3);
+			X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
+			T = _mm_add_epi32(X1, X2);
+			X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
+			X1 = _mm_shuffle_epi32(X1, 0x39);
+			X2 = _mm_shuffle_epi32(X2, 0x4E);
+			X3 = _mm_shuffle_epi32(X3, 0x93);
+		}
 
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
+		X0 = _mm_add_epi32(X0s, X0);
+		X1 = _mm_add_epi32(X1s, X1);
+		X2 = _mm_add_epi32(X2s, X2);
+		X3 = _mm_add_epi32(X3s, X3);
 
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
+		__m128i k02 = _mm_or_si128(_mm_slli_epi64(X0, 32), _mm_srli_epi64(X3, 32));
+		__m128i k20 = _mm_or_si128(_mm_and_si128(X2, maskLo32), _mm_and_si128(X1, maskHi32));
+		__m128i k13 = _mm_or_si128(_mm_slli_epi64(X1, 32), _mm_srli_epi64(X0, 32));
+		__m128i k31 = _mm_or_si128(_mm_and_si128(X3, maskLo32), _mm_and_si128(X2, maskHi32));
+		k02 = _mm_shuffle_epi32(k02, _MM_SHUFFLE(0, 1, 2, 3));
+		k13 = _mm_shuffle_epi32(k13, _MM_SHUFFLE(0, 1, 2, 3));
 
-		X0 = _mm_add_epi32(X0s,X0);
-		X1 = _mm_add_epi32(X1s,X1);
-		X2 = _mm_add_epi32(X2s,X2);
-		X3 = _mm_add_epi32(X3s,X3);
+		_mm_storeu_si128(reinterpret_cast<__m128i *>(c), _mm_xor_si128(_mm_unpackhi_epi64(k02, k20), _mm_loadu_si128(reinterpret_cast<const __m128i *>(m))));
+		_mm_storeu_si128(reinterpret_cast<__m128i *>(c) + 1, _mm_xor_si128(_mm_unpackhi_epi64(k13, k31), _mm_loadu_si128(reinterpret_cast<const __m128i *>(m) + 1)));
+		_mm_storeu_si128(reinterpret_cast<__m128i *>(c) + 2, _mm_xor_si128(_mm_unpacklo_epi64(k20, k02), _mm_loadu_si128(reinterpret_cast<const __m128i *>(m) + 2)));
+		_mm_storeu_si128(reinterpret_cast<__m128i *>(c) + 3, _mm_xor_si128(_mm_unpacklo_epi64(k31, k13), _mm_loadu_si128(reinterpret_cast<const __m128i *>(m) + 3)));
 
-		__m128i k02 = _mm_shuffle_epi32(_mm_or_si128(_mm_slli_epi64(X0, 32), _mm_srli_epi64(X3, 32)), _MM_SHUFFLE(0, 1, 2, 3));
-		__m128i k13 = _mm_shuffle_epi32(_mm_or_si128(_mm_slli_epi64(X1, 32), _mm_srli_epi64(X0, 32)), _MM_SHUFFLE(0, 1, 2, 3));
-		__m128i k20 = _mm_or_si128(_mm_and_si128(X2, _S20SSECONSTANTS.maskLo32), _mm_and_si128(X1, _S20SSECONSTANTS.maskHi32));
-		__m128i k31 = _mm_or_si128(_mm_and_si128(X3, _S20SSECONSTANTS.maskLo32), _mm_and_si128(X2, _S20SSECONSTANTS.maskHi32));
-		_mm_storeu_ps(reinterpret_cast<float *>(c),_mm_castsi128_ps(_mm_xor_si128(_mm_unpackhi_epi64(k02,k20),_mm_castps_si128(_mm_loadu_ps(reinterpret_cast<const float *>(m))))));
-		_mm_storeu_ps(reinterpret_cast<float *>(c) + 4,_mm_castsi128_ps(_mm_xor_si128(_mm_unpackhi_epi64(k13,k31),_mm_castps_si128(_mm_loadu_ps(reinterpret_cast<const float *>(m) + 4)))));
-		_mm_storeu_ps(reinterpret_cast<float *>(c) + 8,_mm_castsi128_ps(_mm_xor_si128(_mm_unpacklo_epi64(k20,k02),_mm_castps_si128(_mm_loadu_ps(reinterpret_cast<const float *>(m) + 8)))));
-		_mm_storeu_ps(reinterpret_cast<float *>(c) + 12,_mm_castsi128_ps(_mm_xor_si128(_mm_unpacklo_epi64(k31,k13),_mm_castps_si128(_mm_loadu_ps(reinterpret_cast<const float *>(m) + 12)))));
+		X0 = X0s;
+		X1 = X1s;
+		X2 = _mm_add_epi32(X2s, add1);
+		X3 = X3s;
 
-		if (!(++_state.i[8])) {
-			++_state.i[5]; // state reordered for SSE
-			/* stopping at 2^70 bytes per nonce is user's responsibility */
-		}
 #else
+
 		x0 = j0;
 		x1 = j1;
 		x2 = j2;
@@ -343,209 +227,40 @@ void Salsa20::crypt12(const void *in,void *out,unsigned int bytes) noexcept
 		x14 = j14;
 		x15 = j15;
 
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
+		for(unsigned int rr=0;rr<(R/2);++rr) {
+			 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
+			 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
+			x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
+			 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
+			 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
+			x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
+			 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
+			 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
+			x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
+			 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
+			 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
+			x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
+			 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
+			 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
+			x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
+			x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
+			 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
+			 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
+			 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
+			 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
+			 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
+			 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
+			 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
+			 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
+			x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
+			 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
+			 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
+			x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
+			x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
+			x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
+			x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
+			x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
+		}
 
 		x0 = PLUS(x0,j0);
 		x1 = PLUS(x1,j1);
@@ -581,761 +296,37 @@ void Salsa20::crypt12(const void *in,void *out,unsigned int bytes) noexcept
 		U32TO8_LITTLE(c + 56,XOR(x14,U8TO32_LITTLE(m + 56)));
 		U32TO8_LITTLE(c + 60,XOR(x15,U8TO32_LITTLE(m + 60)));
 
-		if (!(++j8)) {
-			++j9;
-			/* stopping at 2^70 bytes per nonce is user's responsibility */
-		}
+		++j8;
+
 #endif
 
-		if (bytes <= 64) {
+		if (likely(bytes > 64)) {
+			bytes -= 64;
+			c += 64;
+			m += 64;
+		} else {
 			if (bytes < 64) {
-				for (i = 0;i < bytes;++i)
+				for (unsigned int i = 0;i < bytes;++i)
 					ctarget[i] = c[i];
 			}
-
-#ifndef ZT_SALSA20_SSE
-			_state.i[8] = j8;
-			_state.i[9] = j9;
+#ifdef ZT_SALSA20_SSE
+			state->v[2] = X2;
+#else
+			state->i[8] = j8;
 #endif
-
 			return;
 		}
-
-		bytes -= 64;
-		c += 64;
-		m += 64;
 	}
 }
 
-void Salsa20::crypt20(const void *in,void *out,unsigned int bytes) noexcept
+void Salsa20::crypt12(const void *in, void *out, unsigned int bytes) noexcept
 {
-	uint8_t tmp[64];
-	const uint8_t *m = (const uint8_t *)in;
-	uint8_t *c = (uint8_t *)out;
-	uint8_t *ctarget = c;
-	unsigned int i;
-
-#ifndef ZT_SALSA20_SSE
-	uint32_t x0, x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15;
-	uint32_t j0, j1, j2, j3, j4, j5, j6, j7, j8, j9, j10, j11, j12, j13, j14, j15;
-#endif
-
-	if (!bytes)
-		return;
-
-#ifndef ZT_SALSA20_SSE
-	j0 = _state.i[0];
-	j1 = _state.i[1];
-	j2 = _state.i[2];
-	j3 = _state.i[3];
-	j4 = _state.i[4];
-	j5 = _state.i[5];
-	j6 = _state.i[6];
-	j7 = _state.i[7];
-	j8 = _state.i[8];
-	j9 = _state.i[9];
-	j10 = _state.i[10];
-	j11 = _state.i[11];
-	j12 = _state.i[12];
-	j13 = _state.i[13];
-	j14 = _state.i[14];
-	j15 = _state.i[15];
-#endif
-
-	for (;;) {
-		if (bytes < 64) {
-			for (i = 0;i < bytes;++i)
-				tmp[i] = m[i];
-			m = tmp;
-			ctarget = c;
-			c = tmp;
-		}
-
-#ifdef ZT_SALSA20_SSE
-		__m128i X0 = _mm_loadu_si128((const __m128i *)&(_state.v[0]));
-		__m128i X1 = _mm_loadu_si128((const __m128i *)&(_state.v[1]));
-		__m128i X2 = _mm_loadu_si128((const __m128i *)&(_state.v[2]));
-		__m128i X3 = _mm_loadu_si128((const __m128i *)&(_state.v[3]));
-		__m128i T;
-		__m128i X0s = X0;
-		__m128i X1s = X1;
-		__m128i X2s = X2;
-		__m128i X3s = X3;
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		// 2X round -------------------------------------------------------------
-		T = _mm_add_epi32(X0, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X1, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X3, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x93);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x39);
-		T = _mm_add_epi32(X0, X1);
-		X3 = _mm_xor_si128(_mm_xor_si128(X3, _mm_slli_epi32(T, 7)), _mm_srli_epi32(T, 25));
-		T = _mm_add_epi32(X3, X0);
-		X2 = _mm_xor_si128(_mm_xor_si128(X2, _mm_slli_epi32(T, 9)), _mm_srli_epi32(T, 23));
-		T = _mm_add_epi32(X2, X3);
-		X1 = _mm_xor_si128(_mm_xor_si128(X1, _mm_slli_epi32(T, 13)), _mm_srli_epi32(T, 19));
-		T = _mm_add_epi32(X1, X2);
-		X0 = _mm_xor_si128(_mm_xor_si128(X0, _mm_slli_epi32(T, 18)), _mm_srli_epi32(T, 14));
-		X1 = _mm_shuffle_epi32(X1, 0x39);
-		X2 = _mm_shuffle_epi32(X2, 0x4E);
-		X3 = _mm_shuffle_epi32(X3, 0x93);
-
-		X0 = _mm_add_epi32(X0s,X0);
-		X1 = _mm_add_epi32(X1s,X1);
-		X2 = _mm_add_epi32(X2s,X2);
-		X3 = _mm_add_epi32(X3s,X3);
-
-		__m128i k02 = _mm_shuffle_epi32(_mm_or_si128(_mm_slli_epi64(X0, 32), _mm_srli_epi64(X3, 32)), _MM_SHUFFLE(0, 1, 2, 3));
-		__m128i k13 = _mm_shuffle_epi32(_mm_or_si128(_mm_slli_epi64(X1, 32), _mm_srli_epi64(X0, 32)), _MM_SHUFFLE(0, 1, 2, 3));
-		__m128i k20 = _mm_or_si128(_mm_and_si128(X2, _S20SSECONSTANTS.maskLo32), _mm_and_si128(X1, _S20SSECONSTANTS.maskHi32));
-		__m128i k31 = _mm_or_si128(_mm_and_si128(X3, _S20SSECONSTANTS.maskLo32), _mm_and_si128(X2, _S20SSECONSTANTS.maskHi32));
-		_mm_storeu_ps(reinterpret_cast<float *>(c),_mm_castsi128_ps(_mm_xor_si128(_mm_unpackhi_epi64(k02,k20),_mm_castps_si128(_mm_loadu_ps(reinterpret_cast<const float *>(m))))));
-		_mm_storeu_ps(reinterpret_cast<float *>(c) + 4,_mm_castsi128_ps(_mm_xor_si128(_mm_unpackhi_epi64(k13,k31),_mm_castps_si128(_mm_loadu_ps(reinterpret_cast<const float *>(m) + 4)))));
-		_mm_storeu_ps(reinterpret_cast<float *>(c) + 8,_mm_castsi128_ps(_mm_xor_si128(_mm_unpacklo_epi64(k20,k02),_mm_castps_si128(_mm_loadu_ps(reinterpret_cast<const float *>(m) + 8)))));
-		_mm_storeu_ps(reinterpret_cast<float *>(c) + 12,_mm_castsi128_ps(_mm_xor_si128(_mm_unpacklo_epi64(k31,k13),_mm_castps_si128(_mm_loadu_ps(reinterpret_cast<const float *>(m) + 12)))));
-
-		if (!(++_state.i[8])) {
-			++_state.i[5]; // state reordered for SSE
-			/* stopping at 2^70 bytes per nonce is user's responsibility */
-		}
-#else
-		x0 = j0;
-		x1 = j1;
-		x2 = j2;
-		x3 = j3;
-		x4 = j4;
-		x5 = j5;
-		x6 = j6;
-		x7 = j7;
-		x8 = j8;
-		x9 = j9;
-		x10 = j10;
-		x11 = j11;
-		x12 = j12;
-		x13 = j13;
-		x14 = j14;
-		x15 = j15;
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		// 2X round -------------------------------------------------------------
-		 x4 = XOR( x4,ROTATE(PLUS( x0,x12), 7));
-		 x8 = XOR( x8,ROTATE(PLUS( x4, x0), 9));
-		x12 = XOR(x12,ROTATE(PLUS( x8, x4),13));
-		 x0 = XOR( x0,ROTATE(PLUS(x12, x8),18));
-		 x9 = XOR( x9,ROTATE(PLUS( x5, x1), 7));
-		x13 = XOR(x13,ROTATE(PLUS( x9, x5), 9));
-		 x1 = XOR( x1,ROTATE(PLUS(x13, x9),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x1,x13),18));
-		x14 = XOR(x14,ROTATE(PLUS(x10, x6), 7));
-		 x2 = XOR( x2,ROTATE(PLUS(x14,x10), 9));
-		 x6 = XOR( x6,ROTATE(PLUS( x2,x14),13));
-		x10 = XOR(x10,ROTATE(PLUS( x6, x2),18));
-		 x3 = XOR( x3,ROTATE(PLUS(x15,x11), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x3,x15), 9));
-		x11 = XOR(x11,ROTATE(PLUS( x7, x3),13));
-		x15 = XOR(x15,ROTATE(PLUS(x11, x7),18));
-		 x1 = XOR( x1,ROTATE(PLUS( x0, x3), 7));
-		 x2 = XOR( x2,ROTATE(PLUS( x1, x0), 9));
-		 x3 = XOR( x3,ROTATE(PLUS( x2, x1),13));
-		 x0 = XOR( x0,ROTATE(PLUS( x3, x2),18));
-		 x6 = XOR( x6,ROTATE(PLUS( x5, x4), 7));
-		 x7 = XOR( x7,ROTATE(PLUS( x6, x5), 9));
-		 x4 = XOR( x4,ROTATE(PLUS( x7, x6),13));
-		 x5 = XOR( x5,ROTATE(PLUS( x4, x7),18));
-		x11 = XOR(x11,ROTATE(PLUS(x10, x9), 7));
-		 x8 = XOR( x8,ROTATE(PLUS(x11,x10), 9));
-		 x9 = XOR( x9,ROTATE(PLUS( x8,x11),13));
-		x10 = XOR(x10,ROTATE(PLUS( x9, x8),18));
-		x12 = XOR(x12,ROTATE(PLUS(x15,x14), 7));
-		x13 = XOR(x13,ROTATE(PLUS(x12,x15), 9));
-		x14 = XOR(x14,ROTATE(PLUS(x13,x12),13));
-		x15 = XOR(x15,ROTATE(PLUS(x14,x13),18));
-
-		x0 = PLUS(x0,j0);
-		x1 = PLUS(x1,j1);
-		x2 = PLUS(x2,j2);
-		x3 = PLUS(x3,j3);
-		x4 = PLUS(x4,j4);
-		x5 = PLUS(x5,j5);
-		x6 = PLUS(x6,j6);
-		x7 = PLUS(x7,j7);
-		x8 = PLUS(x8,j8);
-		x9 = PLUS(x9,j9);
-		x10 = PLUS(x10,j10);
-		x11 = PLUS(x11,j11);
-		x12 = PLUS(x12,j12);
-		x13 = PLUS(x13,j13);
-		x14 = PLUS(x14,j14);
-		x15 = PLUS(x15,j15);
-
-		U32TO8_LITTLE(c + 0,XOR(x0,U8TO32_LITTLE(m + 0)));
-		U32TO8_LITTLE(c + 4,XOR(x1,U8TO32_LITTLE(m + 4)));
-		U32TO8_LITTLE(c + 8,XOR(x2,U8TO32_LITTLE(m + 8)));
-		U32TO8_LITTLE(c + 12,XOR(x3,U8TO32_LITTLE(m + 12)));
-		U32TO8_LITTLE(c + 16,XOR(x4,U8TO32_LITTLE(m + 16)));
-		U32TO8_LITTLE(c + 20,XOR(x5,U8TO32_LITTLE(m + 20)));
-		U32TO8_LITTLE(c + 24,XOR(x6,U8TO32_LITTLE(m + 24)));
-		U32TO8_LITTLE(c + 28,XOR(x7,U8TO32_LITTLE(m + 28)));
-		U32TO8_LITTLE(c + 32,XOR(x8,U8TO32_LITTLE(m + 32)));
-		U32TO8_LITTLE(c + 36,XOR(x9,U8TO32_LITTLE(m + 36)));
-		U32TO8_LITTLE(c + 40,XOR(x10,U8TO32_LITTLE(m + 40)));
-		U32TO8_LITTLE(c + 44,XOR(x11,U8TO32_LITTLE(m + 44)));
-		U32TO8_LITTLE(c + 48,XOR(x12,U8TO32_LITTLE(m + 48)));
-		U32TO8_LITTLE(c + 52,XOR(x13,U8TO32_LITTLE(m + 52)));
-		U32TO8_LITTLE(c + 56,XOR(x14,U8TO32_LITTLE(m + 56)));
-		U32TO8_LITTLE(c + 60,XOR(x15,U8TO32_LITTLE(m + 60)));
-
-		if (!(++j8)) {
-			++j9;
-			/* stopping at 2^70 bytes per nonce is user's responsibility */
-		}
-#endif
-
-		if (bytes <= 64) {
-			if (bytes < 64) {
-				for (i = 0;i < bytes;++i)
-					ctarget[i] = c[i];
-			}
-
-#ifndef ZT_SALSA20_SSE
-			_state.i[8] = j8;
-			_state.i[9] = j9;
-#endif
-
-			return;
-		}
+	p_salsaCrypt<12>(reinterpret_cast<p_SalsaState *>(&_state), reinterpret_cast<const uint8_t *>(in), reinterpret_cast<uint8_t *>(out), bytes);
+}
 
-		bytes -= 64;
-		c += 64;
-		m += 64;
-	}
+void Salsa20::crypt20(const void *in, void *out, unsigned int bytes) noexcept
+{
+	p_salsaCrypt<20>(reinterpret_cast<p_SalsaState *>(&_state), reinterpret_cast<const uint8_t *>(in), reinterpret_cast<uint8_t *>(out), bytes);
 }
 
 } // namespace ZeroTier

node/Salsa20.hpp

@@ -28,6 +28,12 @@ namespace ZeroTier {
 
 /**
  * Salsa20 stream cipher
+ *
+ * This supports both the 12-round and 20-round variants.
+ *
+ * SECURITY: this code only support up to 2^32 bytes per key. This is
+ * a minor optimization done here because ZeroTier messages are
+ * nowhere near this large.
  */
 class Salsa20 : public TriviallyCopyable
 {
@@ -39,7 +45,7 @@ public:
 #endif
 
 	ZT_INLINE Salsa20() noexcept {}
-	ZT_INLINE ~Salsa20() { Utils::burn(&_state,sizeof(_state)); }
+	ZT_INLINE ~Salsa20() noexcept { Utils::burn(&_state,sizeof(_state)); }
 
 	/**
 	 * @param key 256-bit (32 byte) key

node/Tests.cpp

@@ -176,7 +176,7 @@ static const C25519TestVector C25519_TEST_VECTORS[ZT_NUM_C25519_TEST_VECTORS] =
 };
 
 #define IDENTITY_V0_KNOWN_GOOD_0 "8e4df28b72:0:ac3d46abe0c21f3cfe7a6c8d6a85cfcffcb82fbd55af6a4d6350657c68200843fa2e16f9418bbd9702cae365f2af5fb4c420908b803a681d4daef6114d78a2d7:bd8dd6e4ce7022d2f812797a80c6ee8ad180dc4ebf301dec8b06d1be08832bddd63a2f1cfa7b2c504474c75bdc8898ba476ef92e8e2d0509f8441985171ff16e"
-#define IDENTITY_V1_KNOWN_GOOD_0 "b0c2badfeb:1:sueysfvujydbkwykbdfemkm5cjgpezjdrzvfczmmfwd2i2ffrrasybhqkz5xegfrrumoidwqyuovprplysmbhtmkim2whjvivub5tcubakzzkhejhqsaiajcu3eooywx3r7sxyflok7b4lgwjv4qqeahkhh4uwog6ke3yqaie2jp3b4wf2pvo2y:xwfmcy2ptfocxnldnkdhzgo4xj73peve3c4ijnlnr442boef7xin34huerixeoes6jsq5g26rvtngjmhqopim7jxssfkw57z2vxidxkutcr4jzu7mmjpnvixwvmbo26nfbd3albf3fyfzi3py6o4bzcnh7thskzvuks5adscqjnseoajjdka"
+#define IDENTITY_V1_KNOWN_GOOD_0 "cb8be88914:1:uhvyeplt7yjunjissek2ndvj6p6aj5jl7fhdgu64dnqld6h6daixc2ngqxm6pu62hxwheioy3jr46qyaxbmy536qvk5asltqqeyuqhhxakzkfyc3ejza52a4qpp6fabitfiuu6337zji47c4exrepurn6blovy5fgocmp7icwkrohqx354kk55a:xw62q5dniqpm4v7tmagukxlzgmegm3gbhx6izatu43vkvgvi6gejcvzlfg3d7ovqduzosawkq7agwx4qriqv56tr57cpdxzhrouuht7thiptbcvkh5yqrsturbw2eiudf4fijl4zhivtpiw4rbcxhkaobzhhapynhnahswppjlpmvnf4ncia"
 
 // --------------------------------------------------------------------------------------------------------------------