diff --git a/sm4/arm64_verify.s b/sm4/arm64_verify.s
deleted file mode 100644
index c6f3dc5..0000000
--- a/sm4/arm64_verify.s
+++ /dev/null
@@ -1,15 +0,0 @@
-#include "textflag.h"
-
-// func tblAsm(in, imm, out *byte)
-TEXT ·tblAsm(SB),NOSPLIT,$0
-  MOVD	in+0(FP), R8
-  MOVD  imm+8(FP), R9
-  MOVD  out+16(FP), R10
-
-  VLD1 (R8), [V0.B16]
-  VLD1 (R9), [V1.B16]
-
-  VTBL V1.B16, [V0.B16], V2.B16
-  VST1 [V2.B16], (R10)
-  RET
-  
\ No newline at end of file
diff --git a/sm4/arm64_verify_test.go b/sm4/arm64_verify_test.go
deleted file mode 100644
index 7c756c8..0000000
--- a/sm4/arm64_verify_test.go
+++ /dev/null
@@ -1,25 +0,0 @@
-//go:build arm64
-// +build arm64
-
-package sm4
-
-import (
-	"reflect"
-	"testing"
-)
-
-//go:noescape
-func tblAsm(in, imm, out *byte)
-
-func TestTblAsm(t *testing.T) {
-	in := []byte{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15}
-	imm := []byte{15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0}
-
-	out := make([]byte, 16)
-	tblAsm(&in[0], &imm[0], &out[0])
-	if !reflect.DeepEqual(out, imm) {
-		t.Errorf("expected=%v, result=%v\n", imm, out)
-	}
-
-	imm = []byte{0, 7, 2, 3}
-}
diff --git a/sm4/asm_arm64.s b/sm4/asm_arm64.s
index bce05ed..afb56a0 100644
--- a/sm4/asm_arm64.s
+++ b/sm4/asm_arm64.s
@@ -114,6 +114,44 @@ GLOBL fk_mask<>(SB), (NOPTR+RODATA), $16
   VEOR XTMP7.B16, y.B16, y.B16;               \
   VEOR x.B16, y.B16, x.B16
 
+#define load_global_data_1() \
+  LDP	flip_mask<>(SB), (R0, R1)    \
+	VMOV	R0, FLIP_MASK.D[0]         \
+	VMOV	R1, FLIP_MASK.D[1]         \
+  LDP	nibble_mask<>(SB), (R0, R1)  \
+	VMOV	R0, NIBBLE_MASK.D[0]       \
+	VMOV	R1, NIBBLE_MASK.D[1]       \
+  LDP	m1_low<>(SB), (R0, R1)       \
+	VMOV	R0, M1L.D[0]               \
+	VMOV	R1, M1L.D[1]               \
+  LDP	m1_high<>(SB), (R0, R1)      \
+	VMOV	R0, M1H.D[0]               \
+	VMOV	R1, M1H.D[1]               \
+  LDP	m2_low<>(SB), (R0, R1)       \
+	VMOV	R0, M2L.D[0]               \
+	VMOV	R1, M2L.D[1]               \
+  LDP	m2_high<>(SB), (R0, R1)      \
+	VMOV	R0, M2H.D[0]               \
+	VMOV	R1, M2H.D[1]               \
+  LDP	fk_mask<>(SB), (R0, R1)      \
+	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]       
+
+#define load_global_data_2() \
+  load_global_data_1()         \
+  LDP	r08_mask<>(SB), (R0, R1) \
+	VMOV	R0, R08_MASK.D[0]      \
+	VMOV	R1, R08_MASK.D[1]      \
+  LDP	r16_mask<>(SB), (R0, R1) \
+	VMOV	R0, R16_MASK.D[0]      \
+	VMOV	R1, R16_MASK.D[1]      \
+  LDP	r24_mask<>(SB), (R0, R1) \
+	VMOV	R0, R24_MASK.D[0]      \
+	VMOV	R1, R24_MASK.D[1]      
+
 // func expandKeyAsm(key *byte, ck, enc, dec *uint32)
 TEXT ·expandKeyAsm(SB),NOSPLIT,$0
   MOVD	key+0(FP), R8
@@ -121,37 +159,7 @@ TEXT ·expandKeyAsm(SB),NOSPLIT,$0
   MOVD  enc+16(FP), R10
   MOVD  dec+24(FP), R11
   
-  LDP	flip_mask<>(SB), (R0, R1)
-	VMOV	R0, FLIP_MASK.D[0]
-	VMOV	R1, FLIP_MASK.D[1]
-  
-  LDP	nibble_mask<>(SB), (R0, R1)
-	VMOV	R0, NIBBLE_MASK.D[0]
-	VMOV	R1, NIBBLE_MASK.D[1]  
-
-  LDP	m1_low<>(SB), (R0, R1)
-	VMOV	R0, M1L.D[0]
-	VMOV	R1, M1L.D[1]  
-
-  LDP	m1_high<>(SB), (R0, R1)
-	VMOV	R0, M1H.D[0]
-	VMOV	R1, M1H.D[1]  
-
-  LDP	m2_low<>(SB), (R0, R1)
-	VMOV	R0, M2L.D[0]
-	VMOV	R1, M2L.D[1]  
-
-  LDP	m2_high<>(SB), (R0, R1)
-	VMOV	R0, M2H.D[0]
-	VMOV	R1, M2H.D[1] 
-
-  LDP	fk_mask<>(SB), (R0, R1)
-	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]
+  load_global_data_1()
 	
   VLD1 (R8), [t0.B16]
   VTBL FLIP_MASK.B16, [t0.B16], t0.B16
@@ -249,49 +257,7 @@ TEXT ·encryptBlocksAsm(SB),NOSPLIT,$0
   VMOV R21, t2.S[3]
   VMOV R22, t3.S[3]    
 
-  LDP	flip_mask<>(SB), (R0, R1)
-	VMOV	R0, FLIP_MASK.D[0]
-	VMOV	R1, FLIP_MASK.D[1]
-  
-  LDP	nibble_mask<>(SB), (R0, R1)
-	VMOV	R0, NIBBLE_MASK.D[0]
-	VMOV	R1, NIBBLE_MASK.D[1]  
-
-  LDP	m1_low<>(SB), (R0, R1)
-	VMOV	R0, M1L.D[0]
-	VMOV	R1, M1L.D[1]  
-
-  LDP	m1_high<>(SB), (R0, R1)
-	VMOV	R0, M1H.D[0]
-	VMOV	R1, M1H.D[1]  
-
-  LDP	m2_low<>(SB), (R0, R1)
-	VMOV	R0, M2L.D[0]
-	VMOV	R1, M2L.D[1]  
-
-  LDP	m2_high<>(SB), (R0, R1)
-	VMOV	R0, M2H.D[0]
-	VMOV	R1, M2H.D[1] 
-
-  LDP	fk_mask<>(SB), (R0, R1)
-	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]
-
-  LDP	r08_mask<>(SB), (R0, R1)
-	VMOV	R0, R08_MASK.D[0]
-	VMOV	R1, R08_MASK.D[1]
-
-  LDP	r16_mask<>(SB), (R0, R1)
-	VMOV	R0, R16_MASK.D[0]
-	VMOV	R1, R16_MASK.D[1]  
-
-  LDP	r24_mask<>(SB), (R0, R1)
-	VMOV	R0, R24_MASK.D[0]
-	VMOV	R1, R24_MASK.D[1]  
+  load_global_data_2()
 
   VTBL FLIP_MASK.B16, [t0.B16], t0.B16
   VTBL FLIP_MASK.B16, [t1.B16], t1.B16
@@ -399,49 +365,7 @@ TEXT ·encryptBlockAsm(SB),NOSPLIT,$0
   VMOV R21, t2.S[0]
   VMOV R22, t3.S[0]
 
-  LDP	flip_mask<>(SB), (R0, R1)
-	VMOV	R0, FLIP_MASK.D[0]
-	VMOV	R1, FLIP_MASK.D[1]
-  
-  LDP	nibble_mask<>(SB), (R0, R1)
-	VMOV	R0, NIBBLE_MASK.D[0]
-	VMOV	R1, NIBBLE_MASK.D[1]  
-
-  LDP	m1_low<>(SB), (R0, R1)
-	VMOV	R0, M1L.D[0]
-	VMOV	R1, M1L.D[1]  
-
-  LDP	m1_high<>(SB), (R0, R1)
-	VMOV	R0, M1H.D[0]
-	VMOV	R1, M1H.D[1]  
-
-  LDP	m2_low<>(SB), (R0, R1)
-	VMOV	R0, M2L.D[0]
-	VMOV	R1, M2L.D[1]  
-
-  LDP	m2_high<>(SB), (R0, R1)
-	VMOV	R0, M2H.D[0]
-	VMOV	R1, M2H.D[1] 
-
-  LDP	fk_mask<>(SB), (R0, R1)
-	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]
-
-  LDP	r08_mask<>(SB), (R0, R1)
-	VMOV	R0, R08_MASK.D[0]
-	VMOV	R1, R08_MASK.D[1]
-
-  LDP	r16_mask<>(SB), (R0, R1)
-	VMOV	R0, R16_MASK.D[0]
-	VMOV	R1, R16_MASK.D[1]  
-
-  LDP	r24_mask<>(SB), (R0, R1)
-	VMOV	R0, R24_MASK.D[0]
-	VMOV	R1, R24_MASK.D[1]  
+  load_global_data_2()
 
   VEOR ZERO.B16, ZERO.B16, ZERO.B16
   EOR R0, R0