diff --git a/sm4/asm_arm64.s b/sm4/asm_arm64.s
index 1647dea..9b4b9b0 100644
--- a/sm4/asm_arm64.s
+++ b/sm4/asm_arm64.s
@@ -158,6 +158,10 @@ TEXT ·expandKeyAsm(SB),NOSPLIT,$0
 	VMOV	R0, FK_MASK.D[0]
 	VMOV	R1, FK_MASK.D[1]
 
+  LDP	inverse_shift_rows<>(SB), (R0, R1)
+	VMOV	R0, INVERSE_SHIFT_ROWS.D[0]
+	VMOV	R1, INVERSE_SHIFT_ROWS.D[1]
+	
   VLD1 (R8), [t0.B16]; 
   VTBL FLIP_MASK.B16, [t0.B16], t0.B16
   VEOR t0.B16, FK_MASK.B16, t0.B16
@@ -176,14 +180,13 @@ TEXT ·expandKeyAsm(SB),NOSPLIT,$0
   VEOR t3.B16, x.B16, x.B16
   VAND x.B16, NIBBLE_MASK.B16, XTMP7.B16
   VTBL XTMP7.B16, [M1L.B16], y.B16
-/*  
   VUSHR $4, x.D2, x.D2
   VAND x.B16, NIBBLE_MASK.B16, XTMP7.B16
   VTBL XTMP7.B16, [M1H.B16], XTMP7.B16
   VEOR y.B16, XTMP7.B16, x.B16
   VTBL INVERSE_SHIFT_ROWS.B16, [x.B16], x.B16
-*/
-  VEOR y.B16, t0.B16, t0.B16
+  
+  VEOR x.B16, t0.B16, t0.B16
   
   VMOV t0.S[0], R2
   MOVW.P R2, 4(R10)