diff --git a/.github/workflows/test_qemu.yml b/.github/workflows/test_qemu.yml
index 3d2aadf..632aff1 100644
--- a/.github/workflows/test_qemu.yml
+++ b/.github/workflows/test_qemu.yml
@@ -30,7 +30,7 @@ jobs:
       uses: actions/checkout@v3
       
     - name: Test
-      run: go test -v -short ./...
+      run: go test -v -short ./cipher/...
       env:
         DISABLE_SM3NI: 1
         DISABLE_SM4NI: 1
diff --git a/cipher/xts_arm64.s b/cipher/xts_arm64.s
new file mode 100644
index 0000000..02a13e7
--- /dev/null
+++ b/cipher/xts_arm64.s
@@ -0,0 +1,150 @@
+//go:build arm64 && !purego
+// +build arm64,!purego
+
+#include "textflag.h"
+
+#define B0 V0
+#define T1 V1
+#define T2 V2
+
+#define POLY V3
+#define ZERO V4
+
+#define tweak R0
+#define GB R1
+#define I R2
+
+// func mul2(tweak *[blockSize]byte, isGB bool)
+TEXT ·mul2(SB),NOSPLIT,$0
+	MOVD tweak+0(FP), tweak
+	MOVB isGB+8(FP), GB
+
+	VLD1 (tweak), [B0.B16]
+
+	VEOR	POLY.B16, POLY.B16, POLY.B16
+	VEOR	ZERO.B16, ZERO.B16, ZERO.B16
+
+	CMP $1, GB
+	BEQ gb_alg
+
+	MOVD	$0x87, I
+	VMOV	I, POLY.D[0]
+
+	VMOV	B0.D[1], I
+	ASR	$63, I
+	VMOV	I, T1.D[0]
+	VMOV	I, T1.D[1]
+	VAND	POLY.B16, T1.B16, T1.B16
+
+	VUSHR	$63, B0.D2, T2.D2
+	VEXT	$8, T2.B16, ZERO.B16, T2.B16
+	VSHL	$1, B0.D2, B0.D2
+	VEOR	T1.B16, B0.B16, B0.B16
+	VEOR	T2.B16, B0.B16, B0.B16
+
+	VST1 [B0.B16], (tweak)
+	RET
+
+gb_alg:
+	MOVD	$0xE1, I
+	LSL	$56, I
+	VMOV	I, POLY.D[1]
+
+	VREV64 B0.B16, B0.B16
+	VEXT	$8, B0.B16, B0.B16, B0.B16
+
+	VMOV	B0.D[0], I
+	LSL $63, I
+	ASR $63, I
+	VMOV	I, T1.D[0]
+	VMOV	I, T1.D[1]
+	VAND	POLY.B16, T1.B16, T1.B16
+
+	VSHL $63, B0.D2, T2.D2
+	VEXT	$8, ZERO.B16, T2.B16, T2.B16
+	VUSHR	$1, B0.D2, B0.D2
+	VEOR	T1.B16, B0.B16, B0.B16
+	VEOR	T2.B16, B0.B16, B0.B16
+
+	VEXT	$8, B0.B16, B0.B16, B0.B16
+	VREV64 B0.B16, B0.B16
+
+	VST1 [B0.B16], (tweak)
+	RET
+
+// func doubleTweaks(tweak *[blockSize]byte, tweaks []byte, isGB bool)
+TEXT ·doubleTweaks(SB),NOSPLIT,$0
+	MOVD tweak+0(FP), tweak
+	MOVD tweaks+8(FP), R3
+	MOVD tweaks_len+16(FP), R4
+	MOVB isGB+32(FP), GB
+
+	LSR $4, R4
+	EOR R5, R5
+
+	VEOR	POLY.B16, POLY.B16, POLY.B16
+	VEOR	ZERO.B16, ZERO.B16, ZERO.B16
+
+	VLD1 (tweak), [B0.B16]
+
+	CMP $1, GB
+	BEQ dt_gb_alg
+
+	MOVD	$0x87, I
+	VMOV	I, POLY.D[0]
+
+loop:
+	VST1.P [B0.B16], 16(R3)
+
+	VMOV	B0.D[1], I
+	ASR	$63, I
+	VMOV	I, T1.D[0]
+	VMOV	I, T1.D[1]
+	VAND	POLY.B16, T1.B16, T1.B16
+
+	VUSHR	$63, B0.D2, T2.D2
+	VEXT	$8, T2.B16, ZERO.B16, T2.B16
+	VSHL	$1, B0.D2, B0.D2
+	VEOR	T1.B16, B0.B16, B0.B16
+	VEOR	T2.B16, B0.B16, B0.B16
+
+	ADD $1, R5
+	CMP R4, R5
+	BNE loop
+
+	VST1 [B0.B16], (tweak)
+	RET
+
+dt_gb_alg:
+	MOVD	$0xE1, I
+	LSL	$56, I
+	VMOV	I, POLY.D[1]
+
+gb_loop:
+	VST1.P [B0.B16], 16(R3)
+
+	VREV64 B0.B16, B0.B16
+	VEXT	$8, B0.B16, B0.B16, B0.B16
+
+	VMOV	B0.D[0], I
+	LSL $63, I
+	ASR $63, I
+	VMOV	I, T1.D[0]
+	VMOV	I, T1.D[1]
+	VAND	POLY.B16, T1.B16, T1.B16
+
+	VSHL $63, B0.D2, T2.D2
+	VEXT	$8, ZERO.B16, T2.B16, T2.B16
+	VUSHR	$1, B0.D2, B0.D2
+	VEOR	T1.B16, B0.B16, B0.B16
+	VEOR	T2.B16, B0.B16, B0.B16
+
+	VEXT	$8, B0.B16, B0.B16, B0.B16
+	VREV64 B0.B16, B0.B16
+
+	ADD $1, R5
+	CMP R4, R5
+	BNE gb_loop
+
+	VST1 [B0.B16], (tweak)	
+	RET
diff --git a/cipher/xts_asm.go b/cipher/xts_asm.go
index 2b824a5..3606e6a 100644
--- a/cipher/xts_asm.go
+++ b/cipher/xts_asm.go
@@ -1,5 +1,5 @@
-//go:build amd64 && !purego
-// +build amd64,!purego
+//go:build (amd64 && !purego) || (arm64 && !purego)
+// +build amd64,!purego arm64,!purego
 
 package cipher
 
diff --git a/cipher/xts_asm_test.go b/cipher/xts_asm_test.go
index d14be04..376bbdd 100644
--- a/cipher/xts_asm_test.go
+++ b/cipher/xts_asm_test.go
@@ -1,5 +1,5 @@
-//go:build amd64 && !purego
-// +build amd64,!purego
+//go:build (amd64 && !purego) || (arm64 && !purego)
+// +build amd64,!purego arm64,!purego
 
 package cipher
 
diff --git a/cipher/xts_generic.go b/cipher/xts_generic.go
index 01c812d..6c17d6a 100644
--- a/cipher/xts_generic.go
+++ b/cipher/xts_generic.go
@@ -1,5 +1,5 @@
-//go:build !amd64 || purego
-// +build !amd64 purego
+//go:build !amd64 && !arm64 || purego
+// +build !amd64,!arm64 purego
 
 package cipher