diff --git a/cipher/ccm.go b/cipher/ccm.go
index 7b26132..bfcfa57 100644
--- a/cipher/ccm.go
+++ b/cipher/ccm.go
@@ -4,12 +4,12 @@ package cipher
 import (
 	goCipher "crypto/cipher"
 	goSubtle "crypto/subtle"
-	"encoding/binary"
 	"math"
 
 	"errors"
 
 	"github.com/emmansun/gmsm/internal/alias"
+	"github.com/emmansun/gmsm/internal/byteorder"
 	"github.com/emmansun/gmsm/internal/subtle"
 )
 
@@ -72,7 +72,6 @@ func NewCCMWithNonceSize(cipher goCipher.Block, size int) (goCipher.AEAD, error)
 // which generates tags with the given length.
 //
 // Tag sizes between 8 and 16 bytes are allowed.
-//
 func NewCCMWithTagSize(cipher goCipher.Block, tagSize int) (goCipher.AEAD, error) {
 	return NewCCMWithNonceAndTagSize(cipher, ccmStandardNonceSize, tagSize)
 }
@@ -133,7 +132,7 @@ func (c *ccm) auth(nonce, plaintext, additionalData []byte, tagMask *[ccmBlockSi
 	}
 	out[0] |= byte(c.tagSize-2) << 2 // M' = ((tagSize - 2) / 2)*8
 	out[0] |= byte(14 - c.nonceSize) // L'
-	binary.BigEndian.PutUint64(out[ccmBlockSize-8:], uint64(len(plaintext)))
+	byteorder.BEPutUint64(out[ccmBlockSize-8:], uint64(len(plaintext)))
 	copy(out[1:], nonce)
 	// B0
 	c.cipher.Encrypt(out[:], out[:])
@@ -143,7 +142,7 @@ func (c *ccm) auth(nonce, plaintext, additionalData []byte, tagMask *[ccmBlockSi
 		// First adata block includes adata length
 		i := 2
 		if n <= 0xfeff { // l(a) < (2^16 - 2^8)
-			binary.BigEndian.PutUint16(block[:i], uint16(n))
+			byteorder.BEPutUint16(block[:i], uint16(n))
 		} else {
 			block[0] = 0xff
 			// If (2^16 - 2^8) <= l(a) < 2^32, then the length field is encoded as
@@ -152,14 +151,14 @@ func (c *ccm) auth(nonce, plaintext, additionalData []byte, tagMask *[ccmBlockSi
 			if n < uint64(1<<32) {
 				block[1] = 0xfe
 				i = 2 + 4
-				binary.BigEndian.PutUint32(block[2:i], uint32(n))
+				byteorder.BEPutUint32(block[2:i], uint32(n))
 			} else {
 				block[1] = 0xff
 				// If 2^32 <= l(a) < 2^64, then the length field is encoded as ten
 				// octets consisting of the octets 0xff, 0xff, and eight octets encoding
 				// l(a) in most-significant-byte-first order.
 				i = 2 + 8
-				binary.BigEndian.PutUint64(block[2:i], uint64(n))
+				byteorder.BEPutUint64(block[2:i], uint64(n))
 			}
 		}
 		i = copy(block[i:], additionalData) // first block start with additional data length
diff --git a/cipher/hctr.go b/cipher/hctr.go
index 03f2b1b..d7f23cf 100644
--- a/cipher/hctr.go
+++ b/cipher/hctr.go
@@ -2,10 +2,10 @@ package cipher
 
 import (
 	_cipher "crypto/cipher"
-	"encoding/binary"
 	"errors"
 
 	"github.com/emmansun/gmsm/internal/alias"
+	"github.com/emmansun/gmsm/internal/byteorder"
 	"github.com/emmansun/gmsm/internal/subtle"
 )
 
@@ -131,8 +131,8 @@ func NewHCTR(cipher _cipher.Block, tweak, hkey []byte) (LengthPreservingMode, er
 	// would expect, say, 4*key to be in index 4 of the table but due to
 	// this bit ordering it will actually be in index 0010 (base 2) = 2.
 	x := hctrFieldElement{
-		binary.BigEndian.Uint64(hkey[:8]),
-		binary.BigEndian.Uint64(hkey[8:blockSize]),
+		byteorder.BEUint64(hkey[:8]),
+		byteorder.BEUint64(hkey[8:blockSize]),
 	}
 	c.productTable[reverseBits(1)] = x
 
@@ -180,8 +180,8 @@ func (h *hctr) mul(y *hctrFieldElement) {
 }
 
 func (h *hctr) updateBlock(block []byte, y *hctrFieldElement) {
-	y.low ^= binary.BigEndian.Uint64(block)
-	y.high ^= binary.BigEndian.Uint64(block[8:])
+	y.low ^= byteorder.BEUint64(block)
+	y.high ^= byteorder.BEUint64(block[8:])
 	h.mul(y)
 }
 
@@ -214,8 +214,8 @@ func (h *hctr) uhash(m []byte, out *[blockSize]byte) {
 	y.high ^= uint64(len(m)+blockSize) * 8
 	h.mul(&y)
 	// output result
-	binary.BigEndian.PutUint64(out[:], y.low)
-	binary.BigEndian.PutUint64(out[8:], y.high)
+	byteorder.BEPutUint64(out[:], y.low)
+	byteorder.BEPutUint64(out[8:], y.high)
 }
 
 func (h *hctr) EncryptBytes(ciphertext, plaintext []byte) {
@@ -281,7 +281,7 @@ func (h *hctr) ctr(dst, src []byte, baseCtr *[blockSize]byte) {
 			for len(src) >= batchSize {
 				for j := 0; j < concCipher.Concurrency(); j++ {
 					// (i)₂
-					binary.BigEndian.PutUint64(num[blockSize-8:], i)
+					byteorder.BEPutUint64(num[blockSize-8:], i)
 					subtle.XORBytes(ctrs[j*blockSize:], baseCtr[:], num)
 					i++
 				}
@@ -295,7 +295,7 @@ func (h *hctr) ctr(dst, src []byte, baseCtr *[blockSize]byte) {
 
 	for len(src) > 0 {
 		// (i)₂
-		binary.BigEndian.PutUint64(num[blockSize-8:], i)
+		byteorder.BEPutUint64(num[blockSize-8:], i)
 		subtle.XORBytes(ctr, baseCtr[:], num)
 		h.cipher.Encrypt(ctr, ctr)
 		n := subtle.XORBytes(dst, src, ctr)
diff --git a/cipher/xts.go b/cipher/xts.go
index 17d0bc0..71b5341 100644
--- a/cipher/xts.go
+++ b/cipher/xts.go
@@ -2,10 +2,10 @@ package cipher
 
 import (
 	_cipher "crypto/cipher"
-	"encoding/binary"
 	"errors"
 
 	"github.com/emmansun/gmsm/internal/alias"
+	"github.com/emmansun/gmsm/internal/byteorder"
 	"github.com/emmansun/gmsm/internal/subtle"
 )
 
@@ -50,7 +50,7 @@ func NewXTSEncrypter(cipherFunc CipherCreator, key, tweakKey, tweak []byte) (_ci
 // block cipher (which must have a block size of 16 bytes) with sector number.
 func NewXTSEncrypterWithSector(cipherFunc CipherCreator, key, tweakKey []byte, sectorNum uint64) (_cipher.BlockMode, error) {
 	tweak := make([]byte, blockSize)
-	binary.LittleEndian.PutUint64(tweak[:8], sectorNum)
+	byteorder.LEPutUint64(tweak[:8], sectorNum)
 	return NewXTSEncrypter(cipherFunc, key, tweakKey, tweak)
 }
 
@@ -66,7 +66,7 @@ func NewGBXTSEncrypter(cipherFunc CipherCreator, key, tweakKey, tweak []byte) (_
 // It follows GB/T 17964-2021.
 func NewGBXTSEncrypterWithSector(cipherFunc CipherCreator, key, tweakKey []byte, sectorNum uint64) (_cipher.BlockMode, error) {
 	tweak := make([]byte, blockSize)
-	binary.LittleEndian.PutUint64(tweak[:8], sectorNum)
+	byteorder.LEPutUint64(tweak[:8], sectorNum)
 	return NewGBXTSEncrypter(cipherFunc, key, tweakKey, tweak)
 }
 
@@ -122,7 +122,7 @@ func NewXTSDecrypter(cipherFunc CipherCreator, key, tweakKey, tweak []byte) (_ci
 // block cipher (which must have a block size of 16 bytes) with sector number for decryption.
 func NewXTSDecrypterWithSector(cipherFunc CipherCreator, key, tweakKey []byte, sectorNum uint64) (_cipher.BlockMode, error) {
 	tweak := make([]byte, blockSize)
-	binary.LittleEndian.PutUint64(tweak[:8], sectorNum)
+	byteorder.LEPutUint64(tweak[:8], sectorNum)
 	return NewXTSDecrypter(cipherFunc, key, tweakKey, tweak)
 }
 
@@ -138,7 +138,7 @@ func NewGBXTSDecrypter(cipherFunc CipherCreator, key, tweakKey, tweak []byte) (_
 // It follows GB/T 17964-2021.
 func NewGBXTSDecrypterWithSector(cipherFunc CipherCreator, key, tweakKey []byte, sectorNum uint64) (_cipher.BlockMode, error) {
 	tweak := make([]byte, blockSize)
-	binary.LittleEndian.PutUint64(tweak[:8], sectorNum)
+	byteorder.LEPutUint64(tweak[:8], sectorNum)
 	return NewGBXTSDecrypter(cipherFunc, key, tweakKey, tweak)
 }
 
@@ -336,7 +336,7 @@ func mul2Generic(tweak *[blockSize]byte, isGB bool) {
 			tweak[0] ^= GF128_FDBK // 1<<7 | 1<<2 | 1<<1 | 1
 		}
 	} else {
-		// GB/T 17964-2021, 
+		// GB/T 17964-2021,
 		// the coefficient of x⁰ can be obtained by tweak[0] >> 7
 		// the coefficient of x⁷ can be obtained by tweak[0] & 1
 		// the coefficient of x¹²⁰ can be obtained by tweak[15] >> 7
diff --git a/drbg/ctr_drbg.go b/drbg/ctr_drbg.go
index e9446fb..e9d26d7 100644
--- a/drbg/ctr_drbg.go
+++ b/drbg/ctr_drbg.go
@@ -2,10 +2,10 @@ package drbg
 
 import (
 	"crypto/cipher"
-	"encoding/binary"
 	"errors"
 	"time"
 
+	"github.com/emmansun/gmsm/internal/byteorder"
 	"github.com/emmansun/gmsm/internal/subtle"
 	"github.com/emmansun/gmsm/sm4"
 )
@@ -185,8 +185,8 @@ func (cd *CtrDrbg) derive(seedMaterial []byte, returnBytes int) []byte {
 
 	// S = counter || len(seed_material) || len(return_bytes) || seed_material || 0x80
 	// len(S) = ((outlen + 4 + 4 + len(seed_material) + 1 + outlen - 1) / outlen) * outlen
-	binary.BigEndian.PutUint32(S[outlen:], uint32(len(seedMaterial)))
-	binary.BigEndian.PutUint32(S[outlen+4:], uint32(returnBytes))
+	byteorder.BEPutUint32(S[outlen:], uint32(len(seedMaterial)))
+	byteorder.BEPutUint32(S[outlen+4:], uint32(returnBytes))
 	copy(S[outlen+8:], seedMaterial)
 	S[outlen+8+len(seedMaterial)] = 0x80
 
@@ -199,7 +199,7 @@ func (cd *CtrDrbg) derive(seedMaterial []byte, returnBytes int) []byte {
 	block := cd.newBlockCipher(key)
 
 	for i := 0; i < blocks; i++ {
-		binary.BigEndian.PutUint32(S, uint32(i))
+		byteorder.BEPutUint32(S, uint32(i))
 		copy(temp[i*outlen:], cd.bcc(block, S))
 	}
 
diff --git a/drbg/hash_drbg.go b/drbg/hash_drbg.go
index 263359d..a8d51da 100644
--- a/drbg/hash_drbg.go
+++ b/drbg/hash_drbg.go
@@ -1,11 +1,11 @@
 package drbg
 
 import (
-	"encoding/binary"
 	"errors"
 	"hash"
 	"time"
 
+	"github.com/emmansun/gmsm/internal/byteorder"
 	"github.com/emmansun/gmsm/sm3"
 )
 
@@ -15,8 +15,8 @@ const HASH_DRBG_MAX_SEED_SIZE = 111
 // HashDrbg hash DRBG structure, its instance is NOT goroutine safe!!!
 type HashDrbg struct {
 	BaseDrbg
-	newHash func() hash.Hash
-	c       []byte
+	newHash  func() hash.Hash
+	c        []byte
 	hashSize int
 }
 
@@ -146,7 +146,7 @@ func (hd *HashDrbg) addH() {
 
 func (hd *HashDrbg) addReseedCounter() {
 	t := make([]byte, hd.seedLength)
-	binary.BigEndian.PutUint64(t[hd.seedLength-8:], hd.reseedCounter)
+	byteorder.BEPutUint64(t[hd.seedLength-8:], hd.reseedCounter)
 	add(t, hd.v, hd.seedLength)
 }
 
@@ -208,7 +208,7 @@ func (hd *HashDrbg) derive(seedMaterial []byte, len int) []byte {
 	md := hd.newHash()
 	limit := uint64(len+hd.hashSize-1) / uint64(hd.hashSize)
 	var requireBytes [4]byte
-	binary.BigEndian.PutUint32(requireBytes[:], uint32(len<<3))
+	byteorder.BEPutUint32(requireBytes[:], uint32(len<<3))
 	var ct byte = 1
 	k := make([]byte, len)
 	for i := 0; i < int(limit); i++ {
diff --git a/internal/bigmod/nat.go b/internal/bigmod/nat.go
index 4f6b7e0..30176da 100644
--- a/internal/bigmod/nat.go
+++ b/internal/bigmod/nat.go
@@ -5,9 +5,10 @@
 package bigmod
 
 import (
-	"encoding/binary"
 	"errors"
 	"math/bits"
+
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 const (
@@ -205,9 +206,9 @@ func (x *Nat) SetOverflowedBytes(b []byte, m *Modulus) *Nat {
 // big-endian encoded uint value.
 func bigEndianUint(buf []byte) uint {
 	if _W == 64 {
-		return uint(binary.BigEndian.Uint64(buf))
+		return uint(byteorder.BEUint64(buf))
 	}
-	return uint(binary.BigEndian.Uint32(buf))
+	return uint(byteorder.BEUint32(buf))
 }
 
 func (x *Nat) setBytes(b []byte) error {
diff --git a/internal/byteorder/byteorder.go b/internal/byteorder/byteorder.go
new file mode 100644
index 0000000..01500a8
--- /dev/null
+++ b/internal/byteorder/byteorder.go
@@ -0,0 +1,149 @@
+// Copyright 2024 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package byteorder provides functions for decoding and encoding
+// little and big endian integer types from/to byte slices.
+package byteorder
+
+func LEUint16(b []byte) uint16 {
+	_ = b[1] // bounds check hint to compiler; see golang.org/issue/14808
+	return uint16(b[0]) | uint16(b[1])<<8
+}
+
+func LEPutUint16(b []byte, v uint16) {
+	_ = b[1] // early bounds check to guarantee safety of writes below
+	b[0] = byte(v)
+	b[1] = byte(v >> 8)
+}
+
+func LEAppendUint16(b []byte, v uint16) []byte {
+	return append(b,
+		byte(v),
+		byte(v>>8),
+	)
+}
+
+func LEUint32(b []byte) uint32 {
+	_ = b[3] // bounds check hint to compiler; see golang.org/issue/14808
+	return uint32(b[0]) | uint32(b[1])<<8 | uint32(b[2])<<16 | uint32(b[3])<<24
+}
+
+func LEPutUint32(b []byte, v uint32) {
+	_ = b[3] // early bounds check to guarantee safety of writes below
+	b[0] = byte(v)
+	b[1] = byte(v >> 8)
+	b[2] = byte(v >> 16)
+	b[3] = byte(v >> 24)
+}
+
+func LEAppendUint32(b []byte, v uint32) []byte {
+	return append(b,
+		byte(v),
+		byte(v>>8),
+		byte(v>>16),
+		byte(v>>24),
+	)
+}
+
+func LEUint64(b []byte) uint64 {
+	_ = b[7] // bounds check hint to compiler; see golang.org/issue/14808
+	return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 |
+		uint64(b[4])<<32 | uint64(b[5])<<40 | uint64(b[6])<<48 | uint64(b[7])<<56
+}
+
+func LEPutUint64(b []byte, v uint64) {
+	_ = b[7] // early bounds check to guarantee safety of writes below
+	b[0] = byte(v)
+	b[1] = byte(v >> 8)
+	b[2] = byte(v >> 16)
+	b[3] = byte(v >> 24)
+	b[4] = byte(v >> 32)
+	b[5] = byte(v >> 40)
+	b[6] = byte(v >> 48)
+	b[7] = byte(v >> 56)
+}
+
+func LEAppendUint64(b []byte, v uint64) []byte {
+	return append(b,
+		byte(v),
+		byte(v>>8),
+		byte(v>>16),
+		byte(v>>24),
+		byte(v>>32),
+		byte(v>>40),
+		byte(v>>48),
+		byte(v>>56),
+	)
+}
+
+func BEUint16(b []byte) uint16 {
+	_ = b[1] // bounds check hint to compiler; see golang.org/issue/14808
+	return uint16(b[1]) | uint16(b[0])<<8
+}
+
+func BEPutUint16(b []byte, v uint16) {
+	_ = b[1] // early bounds check to guarantee safety of writes below
+	b[0] = byte(v >> 8)
+	b[1] = byte(v)
+}
+
+func BEAppendUint16(b []byte, v uint16) []byte {
+	return append(b,
+		byte(v>>8),
+		byte(v),
+	)
+}
+
+func BEUint32(b []byte) uint32 {
+	_ = b[3] // bounds check hint to compiler; see golang.org/issue/14808
+	return uint32(b[3]) | uint32(b[2])<<8 | uint32(b[1])<<16 | uint32(b[0])<<24
+}
+
+func BEPutUint32(b []byte, v uint32) {
+	_ = b[3] // early bounds check to guarantee safety of writes below
+	b[0] = byte(v >> 24)
+	b[1] = byte(v >> 16)
+	b[2] = byte(v >> 8)
+	b[3] = byte(v)
+}
+
+func BEAppendUint32(b []byte, v uint32) []byte {
+	return append(b,
+		byte(v>>24),
+		byte(v>>16),
+		byte(v>>8),
+		byte(v),
+	)
+}
+
+func BEUint64(b []byte) uint64 {
+	_ = b[7] // bounds check hint to compiler; see golang.org/issue/14808
+	return uint64(b[7]) | uint64(b[6])<<8 | uint64(b[5])<<16 | uint64(b[4])<<24 |
+		uint64(b[3])<<32 | uint64(b[2])<<40 | uint64(b[1])<<48 | uint64(b[0])<<56
+}
+
+func BEPutUint64(b []byte, v uint64) {
+	_ = b[7] // early bounds check to guarantee safety of writes below
+	b[0] = byte(v >> 56)
+	b[1] = byte(v >> 48)
+	b[2] = byte(v >> 40)
+	b[3] = byte(v >> 32)
+	b[4] = byte(v >> 24)
+	b[5] = byte(v >> 16)
+	b[6] = byte(v >> 8)
+	b[7] = byte(v)
+}
+
+func BEAppendUint64(b []byte, v uint64) []byte {
+	return append(b,
+		byte(v>>56),
+		byte(v>>48),
+		byte(v>>40),
+		byte(v>>32),
+		byte(v>>24),
+		byte(v>>16),
+		byte(v>>8),
+		byte(v),
+	)
+}
diff --git a/internal/sm2ec/sm2p256_asm.go b/internal/sm2ec/sm2p256_asm.go
index d4b0491..f8d933e 100644
--- a/internal/sm2ec/sm2p256_asm.go
+++ b/internal/sm2ec/sm2p256_asm.go
@@ -18,6 +18,7 @@ import (
 	"runtime"
 	"unsafe"
 
+	"github.com/emmansun/gmsm/internal/byteorder"
 	"golang.org/x/sys/cpu"
 )
 
@@ -385,18 +386,12 @@ var p256Precomputed *[43]p256AffineTable
 //go:embed p256_asm_table.bin
 var p256PrecomputedEmbed string
 
-func leUint64(b []byte) uint64 {
-	_ = b[7] // bounds check hint to compiler; see golang.org/issue/14808
-	return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 |
-		uint64(b[4])<<32 | uint64(b[5])<<40 | uint64(b[6])<<48 | uint64(b[7])<<56
-}
-
 func init() {
 	p256PrecomputedPtr := (*unsafe.Pointer)(unsafe.Pointer(&p256PrecomputedEmbed))
 	if runtime.GOARCH == "s390x" {
 		var newTable [43 * 32 * 2 * 4]uint64
 		for i, x := range (*[43 * 32 * 2 * 4][8]byte)(*p256PrecomputedPtr) {
-			newTable[i] = leUint64(x[:])
+			newTable[i] = byteorder.LEUint64(x[:])
 		}
 		newTablePtr := unsafe.Pointer(&newTable)
 		p256PrecomputedPtr = &newTablePtr
diff --git a/internal/sm2ec/sm2p256_mqv.go b/internal/sm2ec/sm2p256_mqv.go
index e402271..aab6f71 100644
--- a/internal/sm2ec/sm2p256_mqv.go
+++ b/internal/sm2ec/sm2p256_mqv.go
@@ -1,9 +1,10 @@
 package sm2ec
 
 import (
-	"encoding/binary"
 	"errors"
 	"math/bits"
+
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 var p256Order = [4]uint64{0x53bbf40939d54123, 0x7203df6b21c6052b,
@@ -14,20 +15,20 @@ func fromBytes(bytes []byte) (*[4]uint64, error) {
 		return nil, errors.New("invalid scalar length")
 	}
 	var t [4]uint64
-	t[0] = binary.BigEndian.Uint64(bytes[24:])
-	t[1] = binary.BigEndian.Uint64(bytes[16:])
-	t[2] = binary.BigEndian.Uint64(bytes[8:])
-	t[3] = binary.BigEndian.Uint64(bytes)
+	t[0] = byteorder.BEUint64(bytes[24:])
+	t[1] = byteorder.BEUint64(bytes[16:])
+	t[2] = byteorder.BEUint64(bytes[8:])
+	t[3] = byteorder.BEUint64(bytes)
 	return &t, nil
 }
 
 func toBytes(t *[4]uint64) []byte {
 	var bytes [32]byte
 
-	binary.BigEndian.PutUint64(bytes[:], t[3])
-	binary.BigEndian.PutUint64(bytes[8:], t[2])
-	binary.BigEndian.PutUint64(bytes[16:], t[1])
-	binary.BigEndian.PutUint64(bytes[24:], t[0])
+	byteorder.BEPutUint64(bytes[:], t[3])
+	byteorder.BEPutUint64(bytes[8:], t[2])
+	byteorder.BEPutUint64(bytes[16:], t[1])
+	byteorder.BEPutUint64(bytes[24:], t[0])
 
 	return bytes[:]
 }
diff --git a/kdf/kdf.go b/kdf/kdf.go
index 01e0358..df1a3d9 100644
--- a/kdf/kdf.go
+++ b/kdf/kdf.go
@@ -3,8 +3,9 @@ package kdf
 
 import (
 	"encoding"
-	"encoding/binary"
 	"hash"
+
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 // KdfInterface is the interface implemented by some specific Hash implementations.
@@ -30,7 +31,7 @@ func Kdf(newHash func() hash.Hash, z []byte, keyLen int) []byte {
 
 	if marshaler, ok := baseMD.(encoding.BinaryMarshaler); limit == 1 || len(z) < baseMD.BlockSize() || !ok {
 		for i := 0; i < int(limit); i++ {
-			binary.BigEndian.PutUint32(countBytes[:], ct)
+			byteorder.BEPutUint32(countBytes[:], ct)
 			baseMD.Write(z)
 			baseMD.Write(countBytes[:])
 			k = baseMD.Sum(k)
@@ -46,11 +47,11 @@ func Kdf(newHash func() hash.Hash, z []byte, keyLen int) []byte {
 			if err != nil {
 				panic(err)
 			}
-			binary.BigEndian.PutUint32(countBytes[:], ct)
+			byteorder.BEPutUint32(countBytes[:], ct)
 			md.Write(countBytes[:])
 			k = md.Sum(k)
 			ct++
-		}		
+		}
 	}
 
 	return k[:keyLen]
diff --git a/pkcs/internal/md2/md2.go b/pkcs/internal/md2/md2.go
index 376376f..5a98bbd 100644
--- a/pkcs/internal/md2/md2.go
+++ b/pkcs/internal/md2/md2.go
@@ -5,9 +5,10 @@
 package md2
 
 import (
-	"encoding/binary"
 	"errors"
 	"hash"
+
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 // Size the size of a MD2 checksum in bytes.
@@ -95,7 +96,7 @@ func (d *digest) UnmarshalBinary(b []byte) error {
 
 func appendUint64(b []byte, x uint64) []byte {
 	var a [8]byte
-	binary.BigEndian.PutUint64(a[:], x)
+	byteorder.BEPutUint64(a[:], x)
 	return append(b, a[:]...)
 }
 
diff --git a/pkcs/internal/rc2/rc2.go b/pkcs/internal/rc2/rc2.go
index 0e99e4d..89e85e7 100644
--- a/pkcs/internal/rc2/rc2.go
+++ b/pkcs/internal/rc2/rc2.go
@@ -13,10 +13,10 @@ package rc2
 
 import (
 	"crypto/cipher"
-	"encoding/binary"
 	"fmt"
 
 	"github.com/emmansun/gmsm/internal/alias"
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 // The rc2 block size in bytes
@@ -109,10 +109,10 @@ func (c *rc2Cipher) Encrypt(dst, src []byte) {
 		panic("rc2: invalid buffer overlap")
 	}
 
-	r0 := binary.LittleEndian.Uint16(src[0:2])
-	r1 := binary.LittleEndian.Uint16(src[2:4])
-	r2 := binary.LittleEndian.Uint16(src[4:6])
-	r3 := binary.LittleEndian.Uint16(src[6:BlockSize])
+	r0 := byteorder.LEUint16(src[0:2])
+	r1 := byteorder.LEUint16(src[2:4])
+	r2 := byteorder.LEUint16(src[4:6])
+	r3 := byteorder.LEUint16(src[6:BlockSize])
 
 	var j int
 
@@ -197,10 +197,10 @@ func (c *rc2Cipher) Encrypt(dst, src []byte) {
 		j++
 	}
 
-	binary.LittleEndian.PutUint16(dst[0:2], r0)
-	binary.LittleEndian.PutUint16(dst[2:4], r1)
-	binary.LittleEndian.PutUint16(dst[4:6], r2)
-	binary.LittleEndian.PutUint16(dst[6:BlockSize], r3)
+	byteorder.LEPutUint16(dst[0:2], r0)
+	byteorder.LEPutUint16(dst[2:4], r1)
+	byteorder.LEPutUint16(dst[4:6], r2)
+	byteorder.LEPutUint16(dst[6:BlockSize], r3)
 }
 
 func (c *rc2Cipher) Decrypt(dst, src []byte) {
@@ -213,10 +213,10 @@ func (c *rc2Cipher) Decrypt(dst, src []byte) {
 	if alias.InexactOverlap(dst[:BlockSize], src[:BlockSize]) {
 		panic("rc2: invalid buffer overlap")
 	}
-	r0 := binary.LittleEndian.Uint16(src[0:2])
-	r1 := binary.LittleEndian.Uint16(src[2:4])
-	r2 := binary.LittleEndian.Uint16(src[4:6])
-	r3 := binary.LittleEndian.Uint16(src[6:BlockSize])
+	r0 := byteorder.LEUint16(src[0:2])
+	r1 := byteorder.LEUint16(src[2:4])
+	r2 := byteorder.LEUint16(src[4:6])
+	r3 := byteorder.LEUint16(src[6:BlockSize])
 
 	j := 63
 
@@ -301,8 +301,8 @@ func (c *rc2Cipher) Decrypt(dst, src []byte) {
 		j--
 	}
 
-	binary.LittleEndian.PutUint16(dst[0:2], r0)
-	binary.LittleEndian.PutUint16(dst[2:4], r1)
-	binary.LittleEndian.PutUint16(dst[4:6], r2)
-	binary.LittleEndian.PutUint16(dst[6:BlockSize], r3)
+	byteorder.LEPutUint16(dst[0:2], r0)
+	byteorder.LEPutUint16(dst[2:4], r1)
+	byteorder.LEPutUint16(dst[4:6], r2)
+	byteorder.LEPutUint16(dst[6:BlockSize], r3)
 }
diff --git a/sm3/kdf_mult4_asm.go b/sm3/kdf_mult4_asm.go
index 056785c..35bdc32 100644
--- a/sm3/kdf_mult4_asm.go
+++ b/sm3/kdf_mult4_asm.go
@@ -6,7 +6,7 @@
 
 package sm3
 
-import "encoding/binary"
+import "github.com/emmansun/gmsm/internal/byteorder"
 
 // prepare data template: remaining data + [ct] + padding + length
 // p will be 1 or 2 blocks according to the length of remaining data
@@ -18,7 +18,7 @@ func prepareInitData(baseMD *digest, p []byte, len, lenStart uint64) {
 	var tmp [64 + 8]byte // padding + length buffer
 	tmp[0] = 0x80
 	padlen := tmp[:lenStart+8]
-	binary.BigEndian.PutUint64(padlen[lenStart:], len)
+	byteorder.BEPutUint64(padlen[lenStart:], len)
 	copy(p[baseMD.nx+4:], padlen)
 }
 
@@ -34,7 +34,7 @@ func kdfBy4(baseMD *digest, keyLen int, limit int) []byte {
 	if limit < 4 {
 		return kdfGeneric(baseMD, keyLen, limit)
 	}
-	
+
 	var t uint64
 	blocks := 1
 	len := baseMD.len + 4
@@ -55,7 +55,7 @@ func kdfBy4(baseMD *digest, keyLen int, limit int) []byte {
 	var data [parallelSize4][]byte
 	var digs [parallelSize4]*[8]uint32
 	var states [parallelSize4][8]uint32
-	
+
 	for j := 0; j < parallelSize4; j++ {
 		digs[j] = &states[j]
 		p := buffer[blocks*BlockSize*j:]
@@ -77,7 +77,7 @@ func kdfBy4(baseMD *digest, keyLen int, limit int) []byte {
 			// prepare states
 			states[j] = baseMD.h
 			// prepare data
-			binary.BigEndian.PutUint32(data[j][baseMD.nx:], ct)
+			byteorder.BEPutUint32(data[j][baseMD.nx:], ct)
 			ct++
 		}
 		blockMultBy4(&digs[0], &dataPtrs[0], &tmp[0], blocks)
@@ -86,7 +86,7 @@ func kdfBy4(baseMD *digest, keyLen int, limit int) []byte {
 	}
 	remain := limit % parallelSize4
 	for i := 0; i < remain; i++ {
-		binary.BigEndian.PutUint32(tmp[:], ct)
+		byteorder.BEPutUint32(tmp[:], ct)
 		md := *baseMD
 		md.Write(tmp[:4])
 		h := md.checkSum()
diff --git a/sm3/kdf_mult8_amd64.go b/sm3/kdf_mult8_amd64.go
index 1e2fd49..b8b8ae1 100644
--- a/sm3/kdf_mult8_amd64.go
+++ b/sm3/kdf_mult8_amd64.go
@@ -6,7 +6,7 @@
 
 package sm3
 
-import "encoding/binary"
+import "github.com/emmansun/gmsm/internal/byteorder"
 
 // p || state || words
 // p = 64 * 8 * 2 = 1024
@@ -30,7 +30,7 @@ func kdfBy8(baseMD *digest, keyLen int, limit int) []byte {
 	len <<= 3
 
 	var ct uint32 = 1
-	k := make([]byte, limit * Size)
+	k := make([]byte, limit*Size)
 	ret := k
 
 	// prepare temporary buffer
@@ -42,7 +42,7 @@ func kdfBy8(baseMD *digest, keyLen int, limit int) []byte {
 	var data [parallelSize8][]byte
 	var digs [parallelSize8]*[8]uint32
 	var states [parallelSize8][8]uint32
-	
+
 	for j := 0; j < parallelSize8; j++ {
 		digs[j] = &states[j]
 		p := buffer[blocks*BlockSize*j:]
@@ -54,14 +54,14 @@ func kdfBy8(baseMD *digest, keyLen int, limit int) []byte {
 			copy(p, data[0])
 		}
 	}
-	
+
 	times := limit / parallelSize8
 	for i := 0; i < times; i++ {
 		for j := 0; j < parallelSize8; j++ {
 			// prepare states
 			states[j] = baseMD.h
 			// prepare data
-			binary.BigEndian.PutUint32(data[j][baseMD.nx:], ct)
+			byteorder.BEPutUint32(data[j][baseMD.nx:], ct)
 			ct++
 		}
 		blockMultBy8(&digs[0], &dataPtrs[0], &tmp[0], blocks)
@@ -75,7 +75,7 @@ func kdfBy8(baseMD *digest, keyLen int, limit int) []byte {
 			// prepare states
 			states[j] = baseMD.h
 			// prepare data
-			binary.BigEndian.PutUint32(data[j][baseMD.nx:], ct)
+			byteorder.BEPutUint32(data[j][baseMD.nx:], ct)
 			ct++
 		}
 		blockMultBy4(&digs[0], &dataPtrs[0], &tmp[0], blocks)
@@ -85,7 +85,7 @@ func kdfBy8(baseMD *digest, keyLen int, limit int) []byte {
 	}
 
 	for i := 0; i < remain; i++ {
-		binary.BigEndian.PutUint32(tmp[:], ct)
+		byteorder.BEPutUint32(tmp[:], ct)
 		md := *baseMD
 		md.Write(tmp[:4])
 		h := md.checkSum()
diff --git a/sm3/sm3.go b/sm3/sm3.go
index 9be1a78..094cfb2 100644
--- a/sm3/sm3.go
+++ b/sm3/sm3.go
@@ -4,9 +4,10 @@ package sm3
 // [GM/T] SM3 GB/T 32905-2016
 
 import (
-	"encoding/binary"
 	"errors"
 	"hash"
+
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 // Size the size of a SM3 checksum in bytes.
@@ -39,7 +40,7 @@ type digest struct {
 }
 
 const (
-	magic      = "sm3\x03"
+	magic         = "sm3\x03"
 	marshaledSize = len(magic) + 8*4 + chunk + 8
 )
 
@@ -87,13 +88,13 @@ func (d *digest) UnmarshalBinary(b []byte) error {
 
 func appendUint64(b []byte, x uint64) []byte {
 	var a [8]byte
-	binary.BigEndian.PutUint64(a[:], x)
+	byteorder.BEPutUint64(a[:], x)
 	return append(b, a[:]...)
 }
 
 func appendUint32(b []byte, x uint32) []byte {
 	var a [4]byte
-	binary.BigEndian.PutUint32(a[:], x)
+	byteorder.BEPutUint32(a[:], x)
 	return append(b, a[:]...)
 }
 
@@ -143,7 +144,7 @@ func (d *digest) checkSum() [Size]byte {
 	// Length in bits.
 	len <<= 3
 	padlen := tmp[:t+8]
-	binary.BigEndian.PutUint64(padlen[t:], len)
+	byteorder.BEPutUint64(padlen[t:], len)
 	d.Write(padlen)
 
 	if d.nx != 0 {
@@ -152,14 +153,14 @@ func (d *digest) checkSum() [Size]byte {
 
 	var digest [Size]byte
 
-	binary.BigEndian.PutUint32(digest[0:], d.h[0])
-	binary.BigEndian.PutUint32(digest[4:], d.h[1])
-	binary.BigEndian.PutUint32(digest[8:], d.h[2])
-	binary.BigEndian.PutUint32(digest[12:], d.h[3])
-	binary.BigEndian.PutUint32(digest[16:], d.h[4])
-	binary.BigEndian.PutUint32(digest[20:], d.h[5])
-	binary.BigEndian.PutUint32(digest[24:], d.h[6])
-	binary.BigEndian.PutUint32(digest[28:], d.h[7])
+	byteorder.BEPutUint32(digest[0:], d.h[0])
+	byteorder.BEPutUint32(digest[4:], d.h[1])
+	byteorder.BEPutUint32(digest[8:], d.h[2])
+	byteorder.BEPutUint32(digest[12:], d.h[3])
+	byteorder.BEPutUint32(digest[16:], d.h[4])
+	byteorder.BEPutUint32(digest[20:], d.h[5])
+	byteorder.BEPutUint32(digest[24:], d.h[6])
+	byteorder.BEPutUint32(digest[28:], d.h[7])
 
 	return digest
 }
@@ -231,7 +232,7 @@ func kdfGeneric(baseMD *digest, keyLen int, limit int) []byte {
 	var ct uint32 = 1
 	k := make([]byte, keyLen)
 	for i := 0; i < limit; i++ {
-		binary.BigEndian.PutUint32(countBytes[:], ct)
+		byteorder.BEPutUint32(countBytes[:], ct)
 		md := *baseMD
 		md.Write(countBytes[:])
 		h := md.checkSum()
diff --git a/sm3/sm3blocks_test.go b/sm3/sm3blocks_test.go
index 19a0f84..b0209d9 100644
--- a/sm3/sm3blocks_test.go
+++ b/sm3/sm3blocks_test.go
@@ -8,9 +8,10 @@ package sm3
 
 import (
 	"bytes"
-	"encoding/binary"
 	"fmt"
 	"testing"
+
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 func initState4() [4]*[8]uint32 {
@@ -119,7 +120,7 @@ func TestCopyResultsBy4(t *testing.T) {
 
 	for i := 0; i < 4; i++ {
 		for j := 0; j < 8; j++ {
-			binary.BigEndian.PutUint32(expected[i*32+j*4:], m[i][j])
+			byteorder.BEPutUint32(expected[i*32+j*4:], m[i][j])
 		}
 	}
 	if !bytes.Equal(ret[:], expected[:]) {
diff --git a/sm4/block.go b/sm4/block.go
index 18554be..f0e5fb4 100644
--- a/sm4/block.go
+++ b/sm4/block.go
@@ -3,7 +3,7 @@ package sm4
 // [GM/T] SM4 GB/T 32907-2016
 
 import (
-	"encoding/binary"
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 // Encrypt one block from src into dst, using the expanded key xk.
@@ -11,10 +11,10 @@ func encryptBlockGo(xk *[rounds]uint32, dst, src []byte) {
 	_ = src[15] // early bounds check
 
 	var b0, b1, b2, b3 uint32
-	b0 = binary.BigEndian.Uint32(src[0:4])
-	b1 = binary.BigEndian.Uint32(src[4:8])
-	b2 = binary.BigEndian.Uint32(src[8:12])
-	b3 = binary.BigEndian.Uint32(src[12:16])
+	b0 = byteorder.BEUint32(src[0:4])
+	b1 = byteorder.BEUint32(src[4:8])
+	b2 = byteorder.BEUint32(src[8:12])
+	b3 = byteorder.BEUint32(src[12:16])
 
 	// First round uses s-box directly and T transformation.
 	b0 ^= t(b1 ^ b2 ^ b3 ^ xk[0])
@@ -60,10 +60,10 @@ func encryptBlockGo(xk *[rounds]uint32, dst, src []byte) {
 	b3 ^= t(b0 ^ b1 ^ b2 ^ xk[31])
 
 	_ = dst[15] // early bounds check
-	binary.BigEndian.PutUint32(dst[0:4], b3)
-	binary.BigEndian.PutUint32(dst[4:8], b2)
-	binary.BigEndian.PutUint32(dst[8:12], b1)
-	binary.BigEndian.PutUint32(dst[12:16], b0)
+	byteorder.BEPutUint32(dst[0:4], b3)
+	byteorder.BEPutUint32(dst[4:8], b2)
+	byteorder.BEPutUint32(dst[8:12], b1)
+	byteorder.BEPutUint32(dst[12:16], b0)
 }
 
 // Key expansion algorithm.
@@ -71,10 +71,10 @@ func expandKeyGo(key []byte, enc, dec *[rounds]uint32) {
 	// Encryption key setup.
 	key = key[:KeySize]
 	var b0, b1, b2, b3 uint32
-	b0 = binary.BigEndian.Uint32(key[:4]) ^ fk[0]
-	b1 = binary.BigEndian.Uint32(key[4:8]) ^ fk[1]
-	b2 = binary.BigEndian.Uint32(key[8:12]) ^ fk[2]
-	b3 = binary.BigEndian.Uint32(key[12:16]) ^ fk[3]
+	b0 = byteorder.BEUint32(key[:4]) ^ fk[0]
+	b1 = byteorder.BEUint32(key[4:8]) ^ fk[1]
+	b2 = byteorder.BEUint32(key[8:12]) ^ fk[2]
+	b3 = byteorder.BEUint32(key[12:16]) ^ fk[3]
 
 	b0 = b0 ^ t2(b1^b2^b3^ck[0])
 	enc[0], dec[31] = b0, b0
diff --git a/sm4/gcm_cipher_asm.go b/sm4/gcm_cipher_asm.go
index c068019..4a4045f 100644
--- a/sm4/gcm_cipher_asm.go
+++ b/sm4/gcm_cipher_asm.go
@@ -5,10 +5,10 @@ package sm4
 import (
 	"crypto/cipher"
 	goSubtle "crypto/subtle"
-	"encoding/binary"
 	"errors"
 
 	"github.com/emmansun/gmsm/internal/alias"
+	"github.com/emmansun/gmsm/internal/byteorder"
 	"github.com/emmansun/gmsm/internal/subtle"
 )
 
@@ -27,8 +27,8 @@ func (c *sm4CipherAsm) NewGCM(nonceSize, tagSize int) (cipher.AEAD, error) {
 	// would expect, say, 4*key to be in index 4 of the table but due to
 	// this bit ordering it will actually be in index 0010 (base 2) = 2.
 	x := gcmFieldElement{
-		binary.BigEndian.Uint64(key[:8]),
-		binary.BigEndian.Uint64(key[8:]),
+		byteorder.BEUint64(key[:8]),
+		byteorder.BEUint64(key[8:]),
 	}
 	g.productTable[8] = x // reverseBits(1) = 8
 
@@ -48,10 +48,11 @@ func (c *sm4CipherAsm) NewGCM(nonceSize, tagSize int) (cipher.AEAD, error) {
 // gcmFieldElement represents a value in GF(2¹²⁸). In order to reflect the GCM
 // standard and make binary.BigEndian suitable for marshaling these values, the
 // bits are stored in big endian order. For example:
-//   the coefficient of x⁰ can be obtained by v.low >> 63.
-//   the coefficient of x⁶³ can be obtained by v.low & 1.
-//   the coefficient of x⁶⁴ can be obtained by v.high >> 63.
-//   the coefficient of x¹²⁷ can be obtained by v.high & 1.
+//
+//	the coefficient of x⁰ can be obtained by v.low >> 63.
+//	the coefficient of x⁶³ can be obtained by v.low & 1.
+//	the coefficient of x⁶⁴ can be obtained by v.high >> 63.
+//	the coefficient of x¹²⁷ can be obtained by v.high & 1.
 type gcmFieldElement struct {
 	low, high uint64
 }
@@ -233,8 +234,8 @@ func (g *gcm) mul(y *gcmFieldElement) {
 // Horner's rule. There must be a multiple of gcmBlockSize bytes in blocks.
 func (g *gcm) updateBlocks(y *gcmFieldElement, blocks []byte) {
 	for len(blocks) > 0 {
-		y.low ^= binary.BigEndian.Uint64(blocks)
-		y.high ^= binary.BigEndian.Uint64(blocks[8:])
+		y.low ^= byteorder.BEUint64(blocks)
+		y.high ^= byteorder.BEUint64(blocks[8:])
 		g.mul(y)
 		blocks = blocks[gcmBlockSize:]
 	}
@@ -257,7 +258,7 @@ func (g *gcm) update(y *gcmFieldElement, data []byte) {
 // and increments it.
 func gcmInc32(counterBlock *[16]byte) {
 	ctr := counterBlock[len(counterBlock)-4:]
-	binary.BigEndian.PutUint32(ctr, binary.BigEndian.Uint32(ctr)+1)
+	byteorder.BEPutUint32(ctr, byteorder.BEUint32(ctr)+1)
 }
 
 // counterCrypt crypts in to out using g.cipher in counter mode.
@@ -305,8 +306,8 @@ func (g *gcm) deriveCounter(counter *[gcmBlockSize]byte, nonce []byte) {
 		g.update(&y, nonce)
 		y.high ^= uint64(len(nonce)) * 8
 		g.mul(&y)
-		binary.BigEndian.PutUint64(counter[:8], y.low)
-		binary.BigEndian.PutUint64(counter[8:], y.high)
+		byteorder.BEPutUint64(counter[:8], y.low)
+		byteorder.BEPutUint64(counter[8:], y.high)
 	}
 }
 
@@ -322,8 +323,8 @@ func (g *gcm) auth(out, ciphertext, additionalData []byte, tagMask *[gcmTagSize]
 
 	g.mul(&y)
 
-	binary.BigEndian.PutUint64(out, y.low)
-	binary.BigEndian.PutUint64(out[8:], y.high)
+	byteorder.BEPutUint64(out, y.low)
+	byteorder.BEPutUint64(out[8:], y.high)
 
 	subtle.XORBytes(out, out, tagMask[:])
 }
diff --git a/sm4/gcm_ppc64x.go b/sm4/gcm_ppc64x.go
index a2a27e2..5c32cfa 100644
--- a/sm4/gcm_ppc64x.go
+++ b/sm4/gcm_ppc64x.go
@@ -9,11 +9,11 @@ package sm4
 import (
 	"crypto/cipher"
 	_subtle "crypto/subtle"
-	"encoding/binary"
 	"errors"
 	"runtime"
 
 	"github.com/emmansun/gmsm/internal/alias"
+	"github.com/emmansun/gmsm/internal/byteorder"
 	"github.com/emmansun/gmsm/internal/subtle"
 )
 
@@ -60,14 +60,14 @@ func (c *sm4CipherAsm) NewGCM(nonceSize, tagSize int) (cipher.AEAD, error) {
 	// Reverse the bytes in each 8 byte chunk
 	// Load little endian, store big endian
 	if runtime.GOARCH == "ppc64le" {
-		h1 = binary.LittleEndian.Uint64(hle[:8])
-		h2 = binary.LittleEndian.Uint64(hle[8:])
+		h1 = byteorder.LEUint64(hle[:8])
+		h2 = byteorder.LEUint64(hle[8:])
 	} else {
-		h1 = binary.BigEndian.Uint64(hle[:8])
-		h2 = binary.BigEndian.Uint64(hle[8:])
+		h1 = byteorder.BEUint64(hle[:8])
+		h2 = byteorder.BEUint64(hle[8:])
 	}
-	binary.BigEndian.PutUint64(hle[:8], h1)
-	binary.BigEndian.PutUint64(hle[8:], h2)
+	byteorder.BEPutUint64(hle[:8], h1)
+	byteorder.BEPutUint64(hle[8:], h2)
 	gcmInit(&g.productTable, hle)
 
 	return g, nil
@@ -147,8 +147,8 @@ func (g *gcmAsm) counterCrypt(out, in []byte, counter *[gcmBlockSize]byte) {
 // increments the rightmost 32-bits of the count value by 1.
 func gcmInc32(counterBlock *[16]byte) {
 	c := counterBlock[len(counterBlock)-4:]
-	x := binary.BigEndian.Uint32(c) + 1
-	binary.BigEndian.PutUint32(c, x)
+	x := byteorder.BEUint32(c) + 1
+	byteorder.BEPutUint32(c, x)
 }
 
 // paddedGHASH pads data with zeroes until its length is a multiple of
diff --git a/sm4/gcm_ppc64x_test.go b/sm4/gcm_ppc64x_test.go
index b766c0b..8e1b615 100644
--- a/sm4/gcm_ppc64x_test.go
+++ b/sm4/gcm_ppc64x_test.go
@@ -7,10 +7,11 @@
 package sm4
 
 import (
-	"encoding/binary"
 	"fmt"
 	"runtime"
 	"testing"
+
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 func TestCmul(t *testing.T) {
@@ -26,14 +27,14 @@ func TestCmul(t *testing.T) {
 	// Reverse the bytes in each 8 byte chunk
 	// Load little endian, store big endian
 	if runtime.GOARCH == "ppc64le" {
-		h1 = binary.LittleEndian.Uint64(hle[:8])
-		h2 = binary.LittleEndian.Uint64(hle[8:])
+		h1 = byteorder.LEUint64(hle[:8])
+		h2 = byteorder.LEUint64(hle[8:])
 	} else {
-		h1 = binary.BigEndian.Uint64(hle[:8])
-		h2 = binary.BigEndian.Uint64(hle[8:])
+		h1 = byteorder.BEUint64(hle[:8])
+		h2 = byteorder.BEUint64(hle[8:])
 	}
-	binary.BigEndian.PutUint64(hle[:8], h1)
-	binary.BigEndian.PutUint64(hle[8:], h2)
+	byteorder.BEPutUint64(hle[:8], h1)
+	byteorder.BEPutUint64(hle[8:], h2)
 
 	if fmt.Sprintf("%x", hle) != "3811556fff7b1f9fd38f531e5330944d" {
 		t.Errorf("2 got %x", hle)
diff --git a/sm9/bn256/gfp.go b/sm9/bn256/gfp.go
index 92d32ff..b64d1ac 100644
--- a/sm9/bn256/gfp.go
+++ b/sm9/bn256/gfp.go
@@ -1,11 +1,12 @@
 package bn256
 
 import (
-	"encoding/binary"
 	"errors"
 	"fmt"
 	"math/big"
 	"math/bits"
+
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 type gfP [4]uint64
@@ -44,7 +45,7 @@ func fromBigInt(x *big.Int) (out *gfP) {
 	}
 	for i := 0; i < 4; i++ {
 		start := len(bytes) - 8
-		out[i] = binary.BigEndian.Uint64(bytes[start:])
+		out[i] = byteorder.BEUint64(bytes[start:])
 		bytes = bytes[:start]
 	}
 	if x.Sign() < 0 {
diff --git a/sm9/sm9.go b/sm9/sm9.go
index a0f097f..e69a822 100644
--- a/sm9/sm9.go
+++ b/sm9/sm9.go
@@ -4,12 +4,12 @@ package sm9
 import (
 	"crypto"
 	goSubtle "crypto/subtle"
-	"encoding/binary"
 	"errors"
 	"io"
 	"math/big"
 
 	"github.com/emmansun/gmsm/internal/bigmod"
+	"github.com/emmansun/gmsm/internal/byteorder"
 	"github.com/emmansun/gmsm/internal/randutil"
 	"github.com/emmansun/gmsm/internal/subtle"
 	"github.com/emmansun/gmsm/sm3"
@@ -55,7 +55,7 @@ func hash(z []byte, h hashMode) *bigmod.Nat {
 	var countBytes [4]byte
 	var ct uint32 = 1
 
-	binary.BigEndian.PutUint32(countBytes[:], ct)
+	byteorder.BEPutUint32(countBytes[:], ct)
 	md.Write([]byte{byte(h)})
 	md.Write(z)
 	md.Write(countBytes[:])
@@ -63,7 +63,7 @@ func hash(z []byte, h hashMode) *bigmod.Nat {
 	ct++
 	md.Reset()
 
-	binary.BigEndian.PutUint32(countBytes[:], ct)
+	byteorder.BEPutUint32(countBytes[:], ct)
 	md.Write([]byte{byte(h)})
 	md.Write(z)
 	md.Write(countBytes[:])
diff --git a/zuc/core.go b/zuc/core.go
index 148e19b..048e890 100644
--- a/zuc/core.go
+++ b/zuc/core.go
@@ -2,9 +2,10 @@
 package zuc
 
 import (
-	"encoding/binary"
 	"fmt"
 	"math/bits"
+
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 const (
@@ -108,8 +109,8 @@ func (s *zucState32) f32() uint32 {
 	w2 := s.r2 ^ s.x2
 	u := l1((w1 << 16) | (w2 >> 16))
 	v := l2((w2 << 16) | (w1 >> 16))
-	s.r1 = binary.BigEndian.Uint32([]byte{sbox0[u>>24], sbox1[(u>>16)&0xFF], sbox0[(u>>8)&0xFF], sbox1[u&0xFF]})
-	s.r2 = binary.BigEndian.Uint32([]byte{sbox0[v>>24], sbox1[(v>>16)&0xFF], sbox0[(v>>8)&0xFF], sbox1[v&0xFF]})
+	s.r1 = byteorder.BEUint32([]byte{sbox0[u>>24], sbox1[(u>>16)&0xFF], sbox0[(u>>8)&0xFF], sbox1[u&0xFF]})
+	s.r2 = byteorder.BEUint32([]byte{sbox0[v>>24], sbox1[(v>>16)&0xFF], sbox0[(v>>8)&0xFF], sbox1[v&0xFF]})
 	return w
 }
 
diff --git a/zuc/core_test.go b/zuc/core_test.go
index 6a8bd46..06e9dd1 100644
--- a/zuc/core_test.go
+++ b/zuc/core_test.go
@@ -12,7 +12,7 @@ func Test_genKeyword_case1(t *testing.T) {
 		t.Errorf("expected=%x, result=%x\n", 0x27bede74, z1)
 	}
 	if s.r1 != 0xc7ee7f13 {
-		t.Errorf("expected=%x, result=%x\n", 0xc7ee7f13, s.r1)
+		t.Errorf("expected=0xc7ee7f13, result=%x\n", s.r1)
 	}
 	if s.r2 != 0xc0fa817 {
 		t.Errorf("expected=%x, result=%x\n", 0xc0fa817, s.r2)
diff --git a/zuc/eea.go b/zuc/eea.go
index ee189df..726510b 100644
--- a/zuc/eea.go
+++ b/zuc/eea.go
@@ -2,9 +2,9 @@ package zuc
 
 import (
 	"crypto/cipher"
-	"encoding/binary"
 
 	"github.com/emmansun/gmsm/internal/alias"
+	"github.com/emmansun/gmsm/internal/byteorder"
 	"github.com/emmansun/gmsm/internal/subtle"
 )
 
@@ -30,7 +30,7 @@ func NewCipher(key, iv []byte) (cipher.Stream, error) {
 // NewEEACipher create a stream cipher based on key, count, bearer and direction arguments according specification.
 func NewEEACipher(key []byte, count, bearer, direction uint32) (cipher.Stream, error) {
 	iv := make([]byte, 16)
-	binary.BigEndian.PutUint32(iv, count)
+	byteorder.BEPutUint32(iv, count)
 	copy(iv[8:12], iv[:4])
 	iv[4] = byte(((bearer << 1) | (direction & 1)) << 2)
 	iv[12] = iv[4]
@@ -46,7 +46,7 @@ func NewEEACipher(key []byte, count, bearer, direction uint32) (cipher.Stream, e
 func genKeyStreamRev32Generic(keyStream []byte, pState *zucState32) {
 	for len(keyStream) >= 4 {
 		z := genKeyword(pState)
-		binary.BigEndian.PutUint32(keyStream, z)
+		byteorder.BEPutUint32(keyStream, z)
 		keyStream = keyStream[4:]
 	}
 }
diff --git a/zuc/eia.go b/zuc/eia.go
index 1ed2a3b..b2ea401 100644
--- a/zuc/eia.go
+++ b/zuc/eia.go
@@ -1,8 +1,9 @@
 package zuc
 
 import (
-	"encoding/binary"
 	"fmt"
+
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 const (
@@ -60,7 +61,7 @@ func NewHash(key, iv []byte) (*ZUC128Mac, error) {
 
 func genIV4EIA(count, bearer, direction uint32) []byte {
 	iv := make([]byte, 16)
-	binary.BigEndian.PutUint32(iv, count)
+	byteorder.BEPutUint32(iv, count)
 	copy(iv[9:12], iv[1:4])
 	iv[4] = byte(bearer << 3)
 	iv[12] = iv[4]
@@ -102,7 +103,7 @@ func blockGeneric(m *ZUC128Mac, p []byte) {
 		m.genKeywords(m.k0[4:])
 		k64 = uint64(m.k0[0])<<32 | uint64(m.k0[1])
 		// process first 32 bits
-		w := binary.BigEndian.Uint32(p[0:4])
+		w := byteorder.BEUint32(p[0:4])
 		for j := 0; j < 32; j++ {
 			// t64 ^= (w >> 31) ? k64 : 0
 			t64 ^= ^(uint64(w>>31) - 1) & k64
@@ -111,7 +112,7 @@ func blockGeneric(m *ZUC128Mac, p []byte) {
 		}
 		// process second 32 bits
 		k64 = uint64(m.k0[1])<<32 | uint64(m.k0[2])
-		w = binary.BigEndian.Uint32(p[4:8])
+		w = byteorder.BEUint32(p[4:8])
 		for j := 0; j < 32; j++ {
 			t64 ^= ^(uint64(w>>31) - 1) & k64
 			w <<= 1
@@ -119,7 +120,7 @@ func blockGeneric(m *ZUC128Mac, p []byte) {
 		}
 		// process third 32 bits
 		k64 = uint64(m.k0[2])<<32 | uint64(m.k0[3])
-		w = binary.BigEndian.Uint32(p[8:12])
+		w = byteorder.BEUint32(p[8:12])
 		for j := 0; j < 32; j++ {
 			t64 ^= ^(uint64(w>>31) - 1) & k64
 			w <<= 1
@@ -127,7 +128,7 @@ func blockGeneric(m *ZUC128Mac, p []byte) {
 		}
 		// process fourth 32 bits
 		k64 = uint64(m.k0[3])<<32 | uint64(m.k0[4])
-		w = binary.BigEndian.Uint32(p[12:16])
+		w = byteorder.BEUint32(p[12:16])
 		for j := 0; j < 32; j++ {
 			t64 ^= ^(uint64(w>>31) - 1) & k64
 			w <<= 1
@@ -183,7 +184,7 @@ func (m *ZUC128Mac) checkSum(additionalBits int, b byte) [4]byte {
 		// process 32 bits at a time for first complete words
 		for i := 0; i < nwords-1; i++ {
 			k64 = uint64(m.k0[i])<<32 | uint64(m.k0[i+1])
-			w := binary.BigEndian.Uint32(m.x[i*4:])
+			w := byteorder.BEUint32(m.x[i*4:])
 			for j := 0; j < 32; j++ {
 				t64 ^= ^(uint64(w>>31) - 1) & k64
 				w <<= 1
@@ -196,7 +197,7 @@ func (m *ZUC128Mac) checkSum(additionalBits int, b byte) [4]byte {
 		// process remaining bits less than 32
 		if nRemainBits > 0 {
 			k64 = uint64(m.k0[kIdx])<<32 | uint64(m.k0[kIdx+1])
-			w := binary.BigEndian.Uint32(m.x[(nwords-1)*4:])
+			w := byteorder.BEUint32(m.x[(nwords-1)*4:])
 			for j := 0; j < nRemainBits; j++ {
 				t64 ^= ^(uint64(w>>31) - 1) & k64
 				w <<= 1
@@ -212,11 +213,11 @@ func (m *ZUC128Mac) checkSum(additionalBits int, b byte) [4]byte {
 	m.t ^= m.k0[kIdx+1]
 
 	var digest [4]byte
-	binary.BigEndian.PutUint32(digest[:], m.t)
+	byteorder.BEPutUint32(digest[:], m.t)
 	return digest
 }
 
-// Finish this function hash nbits data in p and return mac value, after this function call, 
+// Finish this function hash nbits data in p and return mac value, after this function call,
 // the hash state will be reset.
 // In general, we will use byte level function, this is just for test/verify.
 // nbits: number of bits to hash in p.
diff --git a/zuc/eia256.go b/zuc/eia256.go
index 9d2ba70..b20d784 100644
--- a/zuc/eia256.go
+++ b/zuc/eia256.go
@@ -1,8 +1,9 @@
 package zuc
 
 import (
-	"encoding/binary"
 	"fmt"
+
+	"github.com/emmansun/gmsm/internal/byteorder"
 )
 
 type ZUC256Mac struct {
@@ -93,7 +94,7 @@ func block256Generic(m *ZUC256Mac, p []byte) {
 	for len(p) >= chunk {
 		m.genKeywords(m.k0[4:])
 		for l := 0; l < 4; l++ {
-			w := binary.BigEndian.Uint32(p[l*4:])
+			w := byteorder.BEUint32(p[l*4:])
 			switch m.tagSize {
 			case 4:
 				k64 = uint64(m.k0[l])<<32 | uint64(m.k0[l+1])
@@ -164,7 +165,7 @@ func (m *ZUC256Mac) checkSum(additionalBits int, b byte) []byte {
 		words := (nRemainBits + 31) / 32
 
 		for l := 0; l < words-1; l++ {
-			w := binary.BigEndian.Uint32(m.x[l*4:])
+			w := byteorder.BEUint32(m.x[l*4:])
 			k1 := m.k0[m.tagSize/4+l]
 			for i := 0; i < 32; i++ {
 				wBit := ^(w>>31 - 1)
@@ -183,7 +184,7 @@ func (m *ZUC256Mac) checkSum(additionalBits int, b byte) []byte {
 		nRemainBits -= (words - 1) * 32
 		kIdx = words - 1
 		if nRemainBits > 0 {
-			w := binary.BigEndian.Uint32(m.x[(words-1)*4:])
+			w := byteorder.BEUint32(m.x[(words-1)*4:])
 			for i := 0; i < nRemainBits; i++ {
 				wBit := ^(w>>31 - 1)
 				for j := 0; j < m.tagSize/4; j++ {
@@ -202,7 +203,7 @@ func (m *ZUC256Mac) checkSum(additionalBits int, b byte) []byte {
 	digest := make([]byte, m.tagSize)
 	for j := 0; j < m.tagSize/4; j++ {
 		m.t[j] ^= m.k0[j+kIdx]
-		binary.BigEndian.PutUint32(digest[j*4:], m.t[j])
+		byteorder.BEPutUint32(digest[j*4:], m.t[j])
 	}
 
 	return digest
diff --git a/zuc/eia_test.go b/zuc/eia_test.go
index 7650831..2b40946 100644
--- a/zuc/eia_test.go
+++ b/zuc/eia_test.go
@@ -1,11 +1,11 @@
 package zuc
 
 import (
-	"encoding/binary"
 	"encoding/hex"
 	"hash"
 	"testing"
 
+	"github.com/emmansun/gmsm/internal/byteorder"
 	"github.com/emmansun/gmsm/internal/cryptotest"
 )
 
@@ -144,7 +144,7 @@ func TestEIA_Finish(t *testing.T) {
 		}
 		in := make([]byte, len(test.in)*4)
 		for j, v := range test.in {
-			binary.BigEndian.PutUint32(in[j*4:], v)
+			byteorder.BEPutUint32(in[j*4:], v)
 		}
 
 		mac := h.Finish(in, test.nbits)