diff --git a/sm4/aesni_macros_amd64.s b/sm4/aesni_macros_amd64.s
index eb880ac..85bf876 100644
--- a/sm4/aesni_macros_amd64.s
+++ b/sm4/aesni_macros_amd64.s
@@ -84,60 +84,19 @@ GLOBL fk_mask<>(SB), 8, $16
 //	PUNPCKLQDQ r2, tmp2; 
 //	MOVOU tmp2, r2
 #define SSE_TRANSPOSE_MATRIX(r, r0, r1, r2, r3, tmp1, tmp2) \
-	PEXTRD $2, r0, r; \
-	PINSRD $0, r, tmp2;  \
-	PEXTRD $2, r1, r; \
-	PINSRD $1, r, tmp2;  \
-	; \
-	PEXTRD $3, r0, r; \
-	PINSRD $2, r, tmp2;  \
-	PEXTRD $3, r1, r; \
-	PINSRD $3, r, tmp2;  \ // tmp2 = [w7, w3, w6, w2]
-	; \
-	PEXTRD $1, r0, r; \
-	PINSRD $2, r, r0;  \
-	PEXTRD $0, r1, r; \
-	PINSRD $1, r, r0;  \
-	PEXTRD $1, r1, r; \
-	PINSRD $3, r, r0;  \ //   r0 = [w5, w1, w4, w0] 
-	; \
-	PEXTRD $0, r2, r; \
-	PINSRD $0, r, tmp1;  \
-	PEXTRD $0, r3, r; \
-	PINSRD $1, r, tmp1;  \
-	PEXTRD $1, r2, r; \
-	PINSRD $2, r, tmp1;  \
-	PEXTRD $1, r3, r; \
-	PINSRD $3, r, tmp1;  \ // tmp1 = [w13, w9, w12, w8]
-	; \
-	PEXTRD $2, r2, r; \
-	PINSRD $0, r, r2;  \
-	PEXTRD $2, r3, r; \
-	PINSRD $1, r, r2;  \
-	PEXTRD $3, r2, r; \
-	PINSRD $2, r, r2;  \
-	PEXTRD $3, r3, r; \
-	PINSRD $3, r, r2;  \ //   r2 = [w15, w11, w14, w10] 
-	; \
-	MOVOU r0, r1; \
-	PEXTRQ $1, r1, r; \
-	PINSRQ $0, r, r1; \
-	PEXTRQ $1, tmp1, r; \
-	PINSRQ $1, r, r1; \ //  r1 = [w13, w9, w5, w1]
-	; \
-	PEXTRQ $0, tmp1, r; \
-	PINSRQ $1, r, r0; \ //  r0 = [w12, w8, w4, w0]
-	; \
-	MOVOU tmp2, r3; \
-	PEXTRQ $1, r3, r; \
-	PINSRQ $0, r, r3; \
-	PEXTRQ $1, r2, r; \
-	PINSRQ $1, r, r3; \ //  r3 = [w15, w11, w7, w3]
-	; \
-	PEXTRQ $0, r2, r; \
-	PINSRQ $1, r, r2; \
-	PEXTRQ $0, tmp2, r; \
-	PINSRQ $0, r, r2
+	MOVOU r0, tmp2;      \
+	PUNPCKHLQ r1, tmp2;  \
+	PUNPCKLLQ	r1, r0;  \
+	MOVOU r2, tmp1;      \
+	PUNPCKLLQ r3, tmp1;  \
+	PUNPCKHLQ r3, r2; 	 \
+	MOVOU r0, r1;        \
+	PUNPCKHQDQ tmp1, r1; \
+	PUNPCKLQDQ tmp1, r0; \
+	MOVOU tmp2, r3;      \
+	PUNPCKHQDQ r2, r3;   \
+	PUNPCKLQDQ r2, tmp2; \
+	MOVOU tmp2, r2
 
 // SM4 sbox function
 // parameters:
diff --git a/sm4/asm_amd64.s b/sm4/asm_amd64.s
index 191d554..800c54e 100644
--- a/sm4/asm_amd64.s
+++ b/sm4/asm_amd64.s
@@ -169,29 +169,15 @@ TEXT ·encryptBlocksAsm(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(R12, t0, t1, t2, t3, x, y);  
 
 	XORL CX, CX
 
@@ -205,38 +191,16 @@ 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
-	MOVUPS t3, 0(BX)
-	MOVUPS t2, 16(BX)
-	MOVUPS t1, 32(BX)
-	MOVUPS t0, 48(BX)
-	MOVL  4(BX), R8
-	MOVL  8(BX), R9
-	MOVL  12(BX), R10
-	MOVL  16(BX), R11
-	MOVL  32(BX), R12
-	MOVL  48(BX), R13
-	MOVL  R11, 4(BX)
-	MOVL  R12, 8(BX)
-	MOVL  R13, 12(BX)
-	MOVL  R8, 16(BX)
-	MOVL  R9, 32(BX)
-	MOVL  R10, 48(BX)
-	MOVL  24(BX), R8
-	MOVL  28(BX), R9
-	MOVL  36(BX), R10
-	MOVL  52(BX), R11
-	MOVL  R10, 24(BX)
-	MOVL  R11, 28(BX)
-	MOVL  R8, 36(BX)
-	MOVL  R9, 52(BX)
-	MOVL  44(BX), R8
-	MOVL  56(BX), R9
-	MOVL  R9, 44(BX)
-	MOVL  R8, 56(BX)
+	SSE_TRANSPOSE_MATRIX(R12, 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
+	
+	MOVOU t0, 0(BX)
+	MOVOU t1, 16(BX)
+	MOVOU t2, 32(BX)
+	MOVOU t3, 48(BX)
 
 done_sm4:
 	RET