diff --git a/include/gmssl/gf128.h b/include/gmssl/gf128.h
index 80e535f84..ee325fbb6 100644
--- a/include/gmssl/gf128.h
+++ b/include/gmssl/gf128.h
@@ -1,5 +1,5 @@
 /*
- *  Copyright 2014-2022 The GmSSL Project. All Rights Reserved.
+ *  Copyright 2014-2024 The GmSSL Project. All Rights Reserved.
  *
  *  Licensed under the Apache License, Version 2.0 (the License); you may
  *  not use this file except in compliance with the License.
@@ -27,27 +27,20 @@
 extern "C" {
 #endif
 
-//typedef unsigned __int128 gf128_t;
-
-// the least significant bit of lo is a_0, the most significant bit of hi is a_127
-// so x^7 + x^2 + x + 1 is 0x87
-typedef struct {
-	uint64_t lo;
-	uint64_t hi;
-} gf128_t;
-
-
-// Note: send by value is comptabile with uint128_t and sse2
-gf128_t gf128_from_hex(const char *s);
-int gf128_equ_hex(gf128_t a, const char *s);
-gf128_t gf128_zero(void);
-gf128_t gf128_add(gf128_t a, gf128_t b);
-gf128_t gf128_mul(gf128_t a, gf128_t b);
-gf128_t gf128_mul2(gf128_t a);
-gf128_t gf128_from_bytes(const uint8_t p[16]);
-void gf128_to_bytes(gf128_t a, uint8_t p[16]);
-int gf128_print(FILE *fp, int fmt ,int ind, const char *label, gf128_t a);
-void gf128_print_bits(gf128_t a);
+
+typedef uint64_t gf128_t[2];
+
+void gf128_set_zero(gf128_t r);
+void gf128_set_one(gf128_t r);
+void gf128_add(gf128_t r, const gf128_t a, const gf128_t b);
+void gf128_mul(gf128_t r, const gf128_t a, const gf128_t b);
+void gf128_mul_by_2(gf128_t r, const gf128_t a);
+void gf128_from_bytes(gf128_t r, const uint8_t p[16]);
+void gf128_to_bytes(const gf128_t a, uint8_t p[16]);
+int gf128_from_hex(gf128_t r, const char *s);
+int gf128_equ_hex(const gf128_t a, const char *s);
+int gf128_print(FILE *fp, int fmt, int ind, const char *label, const gf128_t a);
+
 
 #ifdef __cplusplus
 }
diff --git a/src/gf128.c b/src/gf128.c
index c123e6537..515062e13 100644
--- a/src/gf128.c
+++ b/src/gf128.c
@@ -1,5 +1,5 @@
 /*
- *  Copyright 2014-2022 The GmSSL Project. All Rights Reserved.
+ *  Copyright 2014-2024 The GmSSL Project. All Rights Reserved.
  *
  *  Licensed under the Apache License, Version 2.0 (the License); you may
  *  not use this file except in compliance with the License.
@@ -22,6 +22,7 @@
 #include <gmssl/endian.h>
 #include <gmssl/error.h>
 
+
 static uint64_t reverse_bits(uint64_t a)
 {
 	uint64_t r = 0;
@@ -36,30 +37,19 @@ static uint64_t reverse_bits(uint64_t a)
 	return r;
 }
 
-gf128_t gf128_zero(void)
+void gf128_set_zero(gf128_t r)
 {
-	uint8_t zero[16] = {0};
-	return gf128_from_bytes(zero);
+	r[0] = 0;
+	r[1] = 0;
 }
 
-gf128_t gf128_from_hex(const char *s)
+void gf128_set_one(gf128_t r)
 {
-	uint8_t bin[16];
-	size_t len;
-	hex_to_bytes(s, strlen(s), bin, &len);
-	return gf128_from_bytes(bin);
-}
-
-int gf128_equ_hex(gf128_t a, const char *s)
-{
-	uint8_t bin1[16];
-	uint8_t bin2[16];
-	size_t len;
-	hex_to_bytes(s, strlen(s), bin1, &len);
-	gf128_to_bytes(a, bin2);
-	return memcmp(bin1, bin2, sizeof(bin1)) == 0;
+	r[0] = 1;
+	r[1] = 0;
 }
 
+/*
 void gf128_print_bits(gf128_t a)
 {
 	int i;
@@ -77,8 +67,9 @@ void gf128_print_bits(gf128_t a)
 	}
 	printf("\n");
 }
+*/
 
-int gf128_print(FILE *fp, int fmt, int ind, const char *label, gf128_t a)
+int gf128_print(FILE *fp, int fmt, int ind, const char *label, const gf128_t a)
 {
 	uint8_t be[16];
 	int i;
@@ -92,102 +83,117 @@ int gf128_print(FILE *fp, int fmt, int ind, const char *label, gf128_t a)
 	return 1;
 }
 
-gf128_t gf128_from_bytes(const uint8_t p[16])
+void gf128_from_bytes(gf128_t r, const uint8_t p[16])
 {
-	gf128_t r;
-
-	r.lo = GETU64(p);
-	r.hi = GETU64(p + 8);
-
-	r.lo = reverse_bits(r.lo);
-	r.hi = reverse_bits(r.hi);
-	return r;
+	r[0] = reverse_bits(GETU64(p));
+	r[1] = reverse_bits(GETU64(p + 8));
 }
 
-void gf128_to_bytes(gf128_t a, uint8_t p[16])
+void gf128_to_bytes(const gf128_t a, uint8_t p[16])
 {
-	a.lo = reverse_bits(a.lo);
-	a.hi = reverse_bits(a.hi);
-	PUTU64(p, a.lo);
-	PUTU64(p + 8, a.hi);
+	PUTU64(p, reverse_bits(a[0]));
+	PUTU64(p + 8, reverse_bits(a[1]));
 }
 
-gf128_t gf128_add(gf128_t a, gf128_t b)
+void gf128_add(gf128_t r, const gf128_t a, const gf128_t b)
 {
-	gf128_t r;
-	r.hi = a.hi ^ b.hi;
-	r.lo = a.lo ^ b.lo;
-	return r;
+	r[0] = a[0] ^ b[0];
+	r[1] = a[1] ^ b[1];
 }
 
 #ifndef ENABLE_GMUL_AARCH64
-gf128_t gf128_mul(gf128_t a, gf128_t b)
+void gf128_mul(gf128_t r, const gf128_t a, const gf128_t b)
 {
-	gf128_t r = {0, 0};
-	uint64_t mask = (uint64_t)1 << 63;
+	const uint64_t mask = (uint64_t)1 << 63;
+	uint64_t b0 = b[0];
+	uint64_t b1 = b[1];
+	uint64_t r0 = 0; // incase r is a or b
+	uint64_t r1 = 0;
 	int i;
 
+
 	for (i = 0; i < 64; i++) {
-		if (r.hi & mask) {
-			r.hi = r.hi << 1 | r.lo >> 63;
-			r.lo = (r.lo << 1);
-			r.lo ^= 0x87;
+		if (r1 & mask) {
+			r1 = r1 << 1 | r0 >> 63;
+			r0 = r0 << 1;
+			r0 ^= 0x87;
 		} else {
-			r.hi = r.hi << 1 | r.lo >> 63;
-			r.lo = r.lo << 1;
+			r1 = r1 << 1 | r0 >> 63;
+			r0 = r0 << 1;
 		}
 
-		if (b.hi & mask) {
-			r.hi ^= a.hi;
-			r.lo ^= a.lo;
+		if (b1 & mask) {
+			r1 ^= a[1];
+			r0 ^= a[0];
 		}
 
-		b.hi <<= 1;
+		b1 <<= 1;
 	}
 	for (i = 0; i < 64; i++) {
-		if (r.hi & mask) {
-			r.hi = r.hi << 1 | r.lo >> 63;
-			r.lo = (r.lo << 1) ^ 0x87;
+		if (r1 & mask) {
+			r1 = r1 << 1 | r0 >> 63;
+			r0 = r0 << 1;
+			r0 ^= 0x87;
 		} else {
-			r.hi = r.hi << 1 | r.lo >> 63;
-			r.lo = r.lo << 1;
+			r1 = r1 << 1 | r0 >> 63;
+			r0 = r0 << 1;
 		}
 
-		if (b.lo & mask) {
-			r.hi ^= a.hi;
-			r.lo ^= a.lo;
+		if (b0 & mask) {
+			r1 ^= a[1];
+			r0 ^= a[0];
 		}
 
-		b.lo <<= 1;
+		b0 <<= 1;
 	}
 
-	return r;
+	r[0] = r0;
+	r[1] = r1;
 }
-#else
-
-extern void gmul(uint64_t r[2], const uint64_t a[2], const uint64_t b[2]);
+#endif
 
-gf128_t gf128_mul(gf128_t a, gf128_t b)
+void gf128_mul_by_2(gf128_t r, const gf128_t a)
 {
-	gf128_t r;
-	gmul(&r, &a, &b);
-	return r;
-}
+	const uint64_t mask = (uint64_t)1 << 63;
 
-#endif
+	if (a[1] & mask) {
+		r[1] = a[1] << 1 | a[0] >> 63;
+		r[0] = a[0] << 1;
+		r[0] ^= 0x87;
+	} else {
+		r[1] = a[1] << 1 | a[0] >> 63;
+		r[0] = a[0] << 1;
+	}
+}
 
-gf128_t gf128_mul2(gf128_t a)
+int gf128_from_hex(gf128_t r, const char *s)
 {
-	gf128_t r;
+	uint8_t bytes[16];
+	size_t len;
 
-	if (a.hi & ((uint64_t)1 << 63)) {
-		r.hi = a.hi << 1 | a.lo >> 63;
-		r.lo = a.lo << 1;
-		r.lo ^= 0x87;
-	} else {
-		r.hi = a.hi << 1 | a.lo >> 63;
-		r.lo = a.lo << 1;
+	if (strlen(s) != sizeof(bytes) * 2) {
+		error_print();
+		return -1;
 	}
+	if (hex_to_bytes(s, strlen(s), bytes, &len) != 1) {
+		error_print();
+		return -1;
+	}
+	gf128_from_bytes(r, bytes);
+	return 1;
+}
 
-	return r;
+int gf128_equ_hex(const gf128_t a, const char *s)
+{
+	gf128_t b;
+
+	if (gf128_from_hex(b, s) != 1) {
+		error_print();
+		return -1;
+	}
+	if (a[0] != b[0] || a[1] != b[1]) {
+		return 0;
+	}
+	return 1;
 }
+
diff --git a/src/gf128_aarch64.S b/src/gf128_aarch64.S
index ad4d5f634..939e401c3 100644
--- a/src/gf128_aarch64.S
+++ b/src/gf128_aarch64.S
@@ -29,9 +29,9 @@
 	= c + (e0 + w0) * x^64 + (d0 + w1) * f0
 */
 .text
-.globl	_gmul
+.globl	_gf128_mul
 .align	4
-_gmul:
+_gf128_mul:
 	// load (a0, a1)
 	ld1	{v1.2d},[x1]
 	// load (b0, b1)
diff --git a/src/ghash.c b/src/ghash.c
index e67671574..51d4e8cea 100644
--- a/src/ghash.c
+++ b/src/ghash.c
@@ -35,48 +35,51 @@
  */
 void ghash(const uint8_t h[16], const uint8_t *aad, size_t aadlen, const uint8_t *c, size_t clen, uint8_t out[16])
 {
-	gf128_t H = gf128_from_bytes(h);
-	gf128_t X = gf128_zero();
+	gf128_t H;
+	gf128_t X;
 	gf128_t L;
 
+	gf128_from_bytes(H, h);
+	gf128_set_zero(X);
+
 	PUTU64(out, (uint64_t)aadlen << 3);
 	PUTU64(out + 8, (uint64_t)clen << 3);
-	L = gf128_from_bytes(out);
+	gf128_from_bytes(L, out);
 
 	while (aadlen) {
 		gf128_t A;
 		if (aadlen >= 16) {
-			A = gf128_from_bytes(aad);
+			gf128_from_bytes(A, aad);
 			aad += 16;
 			aadlen -= 16;
 		} else {
 			memset(out, 0, 16);
 			memcpy(out, aad, aadlen);
-			A = gf128_from_bytes(out);
+			gf128_from_bytes(A, out);
 			aadlen = 0;
 		}
-		X = gf128_add(X, A);
-		X = gf128_mul(X, H);
+		gf128_add(X, X, A);
+		gf128_mul(X, X, H);
 	}
 
 	while (clen) {
 		gf128_t C;
 		if (clen >= 16) {
-			C = gf128_from_bytes(c);
+			gf128_from_bytes(C, c);
 			c += 16;
 			clen -= 16;
 		} else {
 			memset(out, 0, 16);
 			memcpy(out, c, clen);
-			C = gf128_from_bytes(out);
+			gf128_from_bytes(C, out);
 			clen = 0;
 		}
-		X = gf128_add(X, C);
-		X = gf128_mul(X, H);
+		gf128_add(X, X, C);
+		gf128_mul(X, X, H);
 	}
 
-	X = gf128_add(X, L);
-	H = gf128_mul(X, H);
+	gf128_add(X, X, L);
+	gf128_mul(H, H, X);
 	gf128_to_bytes(H, out);
 }
 
@@ -86,24 +89,24 @@ void ghash_init(GHASH_CTX *ctx, const uint8_t h[16], const uint8_t *aad, size_t
 	gf128_t A;
 
 	memset(ctx, 0, sizeof(*ctx));
-	ctx->H = gf128_from_bytes(h);
-	ctx->X = gf128_zero();
+	gf128_from_bytes(ctx->H, h);
+	gf128_set_zero(ctx->X);
 	ctx->aadlen = aadlen;
 	ctx->clen = 0;
 
 	while (aadlen) {
 		if (aadlen >= 16) {
-			A = gf128_from_bytes(aad);
+			gf128_from_bytes(A, aad);
 			aad += 16;
 			aadlen -= 16;
 		} else {
 			memset(ctx->block, 0, 16);
 			memcpy(ctx->block, aad, aadlen);
-			A = gf128_from_bytes(ctx->block);
+			gf128_from_bytes(A, ctx->block);
 			aadlen = 0;
 		}
-		ctx->X = gf128_add(ctx->X, A);
-		ctx->X = gf128_mul(ctx->X, ctx->H);
+		gf128_add(ctx->X, ctx->X, A);
+		gf128_mul(ctx->X, ctx->X, ctx->H);
 	}
 }
 
@@ -123,18 +126,18 @@ void ghash_update(GHASH_CTX *ctx, const uint8_t *c, size_t clen)
 			return;
 		} else {
 			memcpy(ctx->block + ctx->num, c, left);
-			C = gf128_from_bytes(ctx->block);
-			ctx->X = gf128_add(ctx->X, C);
-			ctx->X = gf128_mul(ctx->X, ctx->H);
+			gf128_from_bytes(C, ctx->block);
+			gf128_add(ctx->X, ctx->X, C);
+			gf128_mul(ctx->X, ctx->X, ctx->H);
 			c += left;
 			clen -= left;
 		}
 	}
 
 	while (clen >= 16) {
-		C = gf128_from_bytes(c);
-		ctx->X = gf128_add(ctx->X, C);
-		ctx->X = gf128_mul(ctx->X, ctx->H);
+		gf128_from_bytes(C, c);
+		gf128_add(ctx->X, ctx->X, C);
+		gf128_mul(ctx->X, ctx->X, ctx->H);
 		c += 16;
 		clen -= 16;
 	}
@@ -152,17 +155,17 @@ void ghash_finish(GHASH_CTX *ctx, uint8_t out[16])
 
 	if (ctx->num) {
 		memset(ctx->block + ctx->num, 0, 16 - ctx->num);
-		C = gf128_from_bytes(ctx->block);
-		ctx->X = gf128_add(ctx->X, C);
-		ctx->X = gf128_mul(ctx->X, ctx->H);
+		gf128_from_bytes(C, ctx->block);
+		gf128_add(ctx->X, ctx->X, C);
+		gf128_mul(ctx->X, ctx->X, ctx->H);
 	}
 
 	PUTU64(ctx->block, (uint64_t)ctx->aadlen << 3);
 	PUTU64(ctx->block + 8, (uint64_t)ctx->clen << 3);
-	L = gf128_from_bytes(ctx->block);
+	gf128_from_bytes(L, ctx->block);
 
-	ctx->X = gf128_add(ctx->X, L);
-	ctx->H = gf128_mul(ctx->X, ctx->H);
+	gf128_add(ctx->X, ctx->X, L);
+	gf128_mul(ctx->H, ctx->X, ctx->H);
 	gf128_to_bytes(ctx->H, out);
 
 	gmssl_secure_clear(ctx, sizeof(*ctx));
diff --git a/tests/gf128test.c b/tests/gf128test.c
index 47a99ead4..ddc298c0e 100644
--- a/tests/gf128test.c
+++ b/tests/gf128test.c
@@ -77,25 +77,17 @@ int test_gf128_mul_more(void)
 
 		printf("test %zu\n", i);
 
-		a = gf128_from_hex(tests[i].a);
-		b = gf128_from_hex(tests[i].b);
+		gf128_from_hex(a, tests[i].a);
+		gf128_from_hex(b, tests[i].b);
 
-		printf("a0 = %llx, a1 = %llx\n", a.lo,  a.hi);
-		printf("b0 = %llx, b1 = %llx\n", b.lo,  b.hi);
+		gf128_mul(r, a, b);
 
-		r = gf128_mul(a, b);
-
-		printf("r0 = %llx, r1 = %llx\n", r.lo,  r.hi);
-			gf128_print_bits(a);
-			gf128_print_bits(b);
-			gf128_print_bits(r);
+		if (gf128_equ_hex(r, tests[i].r) != 1) {
 
-			gf128_to_bytes(r, buf);
-			format_bytes(stderr, 0, 0, "r" ,buf ,16);
 
-		if (gf128_equ_hex(r, tests[i].r) != 1) {
+			printf("error: %s\n", tests[i].label);
 			error_print();
-			//return -1;
+			return -1;
 		}
 	}
 
@@ -103,39 +95,6 @@ int test_gf128_mul_more(void)
 	return 1;
 }
 
-int test_gf128_armv8(void)
-{
-	gf128_t a = { 1, 0 };
-	gf128_t b = { 1, 0 };
-
-	a = gf128_from_hex("de300f9301a499a965f8bf677e99e80d");
-	b = gf128_from_hex("14b267838ec9ef1bb7b5ce8c19e34bc6");
-
-	// pmull 是对低位做了乘法
-	//gf128_print_bits(b);
-
-	gf128_t c = gf128_mul(a, b);
-
-	gf128_print_bits(c);
-	return 1;
-}
-
-
-int test_gf128_print(void)
-{
-	gf128_t a = { 0, 0x8000000000000000 }; // a = 1 + 0*x + ... + 0*x^127
-	gf128_print(stderr, 0, 0, "1 + 0*x + ... + 0*x^127", a);
-
-	// 这个函数打印的不对，因为真正的值是需要 reverse_bits 的，但是这里我们没有反转
-	gf128_print_bits(a);
-
-	// 看来这个比较奇怪了
-
-
-	return 1;
-
-}
-
 int test_gf128_from_hex(void)
 {
 	char *tests[] = {
@@ -149,7 +108,7 @@ int test_gf128_from_hex(void)
 	int i;
 
 	for (i = 0; i < sizeof(tests)/sizeof(tests[0]); i++) {
-		a = gf128_from_hex(tests[i]);
+		gf128_from_hex(a, tests[i]);
 		if (gf128_equ_hex(a, tests[i]) != 1) {
 			error_print();
 			return -1;
@@ -160,7 +119,7 @@ int test_gf128_from_hex(void)
 	return 1;
 }
 
-int test_gf128_mul2(void)
+int test_gf128_mul_by_2(void)
 {
 	char *tests[] = {
 		"00000000000000000000000000000001",
@@ -174,8 +133,8 @@ int test_gf128_mul2(void)
 	int i;
 
 	for (i = 0; i < sizeof(tests)/sizeof(tests[0]); i++) {
-		a = gf128_from_hex(tests[i]);
-		a = gf128_mul2(a);
+		gf128_from_hex(a, tests[i]);
+		gf128_mul_by_2(a, a);
 		if (gf128_equ_hex(a, results[i]) != 1) {
 			error_print();
 			return -1;
@@ -196,16 +155,16 @@ int test_gf128_mul(void)
 	char *hex_mul_a_b = "7d87dda57a20b0c51d9743071ab14010";
 	gf128_t a, b, r;
 
-	a = gf128_from_hex(hex_a);
-	b = gf128_from_hex(hex_b);
+	gf128_from_hex(a, hex_a);
+	gf128_from_hex(b, hex_b);
 
-	r = gf128_add(a, b);
+	gf128_add(r, a, b);
 	if (gf128_equ_hex(r, hex_add_a_b) != 1) {
 		error_print();
 		return -1;
 	}
 
-	r = gf128_mul(a, b);
+	gf128_mul(r, a, b);
 	if (gf128_equ_hex(r, hex_mul_a_b) != 1) {
 		error_print();
 		return -1;
@@ -217,11 +176,9 @@ int test_gf128_mul(void)
 
 int main(void)
 {
-	if (test_gf128_armv8() != 1) goto err;
+	if (test_gf128_mul_by_2() != 1) goto err;
 	if (test_gf128_mul_more() != 1) goto err;
-	if (test_gf128_print() != 1) goto err;
 	if (test_gf128_from_hex() != 1) goto err;
-	if (test_gf128_mul2() != 1) goto err;
 	if (test_gf128_mul() != 1) goto err;
 	printf("%s all tests passed\n", __FILE__);
 	return 0;