From 74d4a07adeb79460caa5a14cb5d3d8e12c1d90ff Mon Sep 17 00:00:00 2001
From: Zhi Guan <guan@pku.edu.cn>
Date: Fri, 6 Jul 2018 23:07:32 +0800
Subject: [PATCH] optimize zuc

about 15% faster on 64-bit cpu
---
 crypto/zuc/zuc_core.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/crypto/zuc/zuc_core.c b/crypto/zuc/zuc_core.c
index a4cc8367..244667f6 100644
--- a/crypto/zuc/zuc_core.c
+++ b/crypto/zuc/zuc_core.c
@@ -124,6 +124,24 @@ static const uint8_t S1[256] = {
 	{int j; for (j=0; j<15;j++) LFSR[j]=LFSR[j+1];}	\
 	LFSR[15] = V
 
+/* FIXME: check if uint64_t is supported */
+#if 1
+#define LFSRWithWorkMode()				\
+	{						\
+		int j;					\
+		uint64_t a = LFSR[0];			\
+		a += ((uint64_t)LFSR[0]) << 8;		\
+		a += ((uint64_t)LFSR[4]) << 20;		\
+		a += ((uint64_t)LFSR[10]) << 21;	\
+		a += ((uint64_t)LFSR[13]) << 17;	\
+		a += ((uint64_t)LFSR[15]) << 15;	\
+		a = (a & 0x7fffffff) + (a >> 31);	\
+		a = (a & 0x7fffffff) + (a >> 31);	\
+		for (j = 0; j < 15; j++)		\
+			LFSR[j] = LFSR[j+1];		\
+		LFSR[15] = a;				\
+	}
+#else
 #define LFSRWithWorkMode()				\
 	V = LFSR[0];					\
 	ADD31(V, ROT31(LFSR[0], 8));			\
@@ -133,6 +151,7 @@ static const uint8_t S1[256] = {
 	ADD31(V, ROT31(LFSR[15], 15));			\
 	{int j; for (j=0; j<15;j++) LFSR[j]=LFSR[j+1];}	\
 	LFSR[15] = V
+#endif
 
 #define BitReconstruction2(X1,X2)					\
 	X1 = ((LFSR[11] & 0xFFFF) << 16) | (LFSR[9] >> 15);		\
@@ -176,6 +195,7 @@ static const uint8_t S1[256] = {
 	(X0 ^ R1) + R2;					\
 	F_(X1, X2)
 
+
 void ZUC_set_key(ZUC_KEY *key, const unsigned char *user_key, const unsigned char *iv)
 {
 	ZUC_UINT31 *LFSR = key->LFSR;