diff --git a/sm4/aesni_macros_amd64.s b/sm4/aesni_macros_amd64.s
index 85bf876..10cb9a8 100644
--- a/sm4/aesni_macros_amd64.s
+++ b/sm4/aesni_macros_amd64.s
@@ -83,13 +83,13 @@ GLOBL fk_mask<>(SB), 8, $16
 //	PUNPCKHQDQ r2, r3; 
 //	PUNPCKLQDQ r2, tmp2; 
 //	MOVOU tmp2, r2
-#define SSE_TRANSPOSE_MATRIX(r, r0, r1, r2, r3, tmp1, tmp2) \
+#define SSE_TRANSPOSE_MATRIX(r0, r1, r2, r3, tmp1, tmp2) \
 	MOVOU r0, tmp2;      \
 	PUNPCKHLQ r1, tmp2;  \
 	PUNPCKLLQ	r1, r0;  \
 	MOVOU r2, tmp1;      \
 	PUNPCKLLQ r3, tmp1;  \
-	PUNPCKHLQ r3, r2; 	 \
+	PUNPCKHLQ r3, r2;    \
 	MOVOU r0, r1;        \
 	PUNPCKHQDQ tmp1, r1; \
 	PUNPCKLQDQ tmp1, r0; \
diff --git a/sm4/asm_amd64.s b/sm4/asm_amd64.s
index 800c54e..84d4f0c 100644
--- a/sm4/asm_amd64.s
+++ b/sm4/asm_amd64.s
@@ -177,7 +177,7 @@ non_avx2_start:
 	PSHUFB flip_mask<>(SB), t1
 	PSHUFB flip_mask<>(SB), t2
 	PSHUFB flip_mask<>(SB), t3
-	SSE_TRANSPOSE_MATRIX(R12, t0, t1, t2, t3, x, y);  
+	SSE_TRANSPOSE_MATRIX(t0, t1, t2, t3, x, y)
 
 	XORL CX, CX
 
@@ -191,7 +191,7 @@ loop:
 		CMPL CX, $4*32
 		JB loop
 
-	SSE_TRANSPOSE_MATRIX(R12, t0, t1, t2, t3, x, y);  
+	SSE_TRANSPOSE_MATRIX(t0, t1, t2, t3, x, y);  
 	PSHUFB bswap_mask<>(SB), t3
 	PSHUFB bswap_mask<>(SB), t2
 	PSHUFB bswap_mask<>(SB), t1
diff --git a/sm4/cbc_cipher_asm_amd64.s b/sm4/cbc_cipher_asm_amd64.s
index 6b38e3d..e46e232 100644
--- a/sm4/cbc_cipher_asm_amd64.s
+++ b/sm4/cbc_cipher_asm_amd64.s
@@ -156,29 +156,15 @@ TEXT ·decryptBlocksChain(SB),NOSPLIT,$0
 	JE   avx
 
 non_avx2_start:
-	PINSRD $0, 0(DX), t0
-	PINSRD $1, 16(DX), t0
-	PINSRD $2, 32(DX), t0
-	PINSRD $3, 48(DX), t0
+	MOVOU 0(DX), t0
+	MOVOU 16(DX), t1
+	MOVOU 32(DX), t2
+	MOVOU 48(DX), t3
 	PSHUFB flip_mask<>(SB), t0
-
-	PINSRD $0, 4(DX), t1
-	PINSRD $1, 20(DX), t1
-	PINSRD $2, 36(DX), t1
-	PINSRD $3, 52(DX), t1
 	PSHUFB flip_mask<>(SB), t1
-
-	PINSRD $0, 8(DX), t2
-	PINSRD $1, 24(DX), t2
-	PINSRD $2, 40(DX), t2
-	PINSRD $3, 56(DX), t2
 	PSHUFB flip_mask<>(SB), t2
-
-	PINSRD $0, 12(DX), t3
-	PINSRD $1, 28(DX), t3
-	PINSRD $2, 44(DX), t3
-	PINSRD $3, 60(DX), t3
 	PSHUFB flip_mask<>(SB), t3
+	SSE_TRANSPOSE_MATRIX(t0, t1, t2, t3, x, y)
 
 	XORL CX, CX
 
@@ -192,22 +178,21 @@ loop:
 		CMPL CX, $4*32
 		JB loop
 
-	PSHUFB flip_mask<>(SB), t3
-	PSHUFB flip_mask<>(SB), t2
-	PSHUFB flip_mask<>(SB), t1
-	PSHUFB flip_mask<>(SB), t0
+	SSE_TRANSPOSE_MATRIX(t0, t1, t2, t3, x, y);  
+	PSHUFB bswap_mask<>(SB), t3
+	PSHUFB bswap_mask<>(SB), t2
+	PSHUFB bswap_mask<>(SB), t1
+	PSHUFB bswap_mask<>(SB), t0
 
-	SSE_TRANSPOSE_MATRIX(CX, t3, t2, t1, t0, XWORD, YWORD)
+	PXOR 0(SI), t0
+	PXOR 16(SI), t1
+	PXOR 32(SI), t2
+	PXOR 48(SI), t3
 
-	PXOR 0(SI), t3
-	PXOR 16(SI), t2
-	PXOR 32(SI), t1
-	PXOR 48(SI), t0
-
-	MOVUPS t3, 0(BX)
-	MOVUPS t2, 16(BX)
-	MOVUPS t1, 32(BX)
-	MOVUPS t0, 48(BX)
+	MOVUPS t0, 0(BX)
+	MOVUPS t1, 16(BX)
+	MOVUPS t2, 32(BX)
+	MOVUPS t3, 48(BX)
 
 done_sm4:
 	RET
diff --git a/sm4/gcm_amd64.s b/sm4/gcm_amd64.s
index d964838..4823fdf 100644
--- a/sm4/gcm_amd64.s
+++ b/sm4/gcm_amd64.s
@@ -216,7 +216,7 @@ TEXT ·gcmSm4Finish(SB),NOSPLIT,$0
 	PSHUFB flip_mask<>(SB), t1; \
 	PSHUFB flip_mask<>(SB), t2; \
 	PSHUFB flip_mask<>(SB), t3; \
-	SSE_TRANSPOSE_MATRIX(R12, t0, t1, t2, t3, x, y);          \
+	SSE_TRANSPOSE_MATRIX(t0, t1, t2, t3, x, y);          \
 	XORL IND, IND;                                            \
 	SM4_ROUND(0, RK, IND, x, y, z, t0, t1, t2, t3);           \
 	SM4_ROUND(1, RK, IND, x, y, z, t1, t2, t3, t0);           \
@@ -257,7 +257,7 @@ TEXT ·gcmSm4Finish(SB),NOSPLIT,$0
 	SM4_ROUND(1, RK, IND, x, y, z, t1, t2, t3, t0);           \
 	SM4_ROUND(2, RK, IND, x, y, z, t2, t3, t0, t1);           \
 	SM4_ROUND(3, RK, IND, x, y, z, t3, t0, t1, t2);           \
-	SSE_TRANSPOSE_MATRIX(R12, t0, t1, t2, t3, x, y);          \
+	SSE_TRANSPOSE_MATRIX(t0, t1, t2, t3, x, y);          \
 	PSHUFB BSWAP, t3; \
 	PSHUFB BSWAP, t2; \
 	PSHUFB BSWAP, t1; \