crypto/cipher: optimize safeXORBytes

Optimize safeXORBytes based on this conversation: #35381

Author: tbeeck

src/crypto/cipher/xor_generic.go

@@ -7,6 +7,7 @@
 package cipher
 
 import (
+	"encoding/binary"
 	"runtime"
 	"unsafe"
 )
@@ -63,7 +64,19 @@ func fastXORBytes(dst, a, b []byte, n int) {
 
 // n needs to be smaller or equal than the length of a and b.
 func safeXORBytes(dst, a, b []byte, n int) {
-	for i := 0; i < n; i++ {
+	// Load multiple bytes so the compiler can recognize
+	// and optimize them into single multi-byte loads
+	w := n / wordSize
+	if w > 0 {
+		for i := 0; i < w; i++ {
+			offset := i * wordSize
+			k := binary.LittleEndian.Uint64(a[offset:])
+			v := binary.LittleEndian.Uint64(b[offset:])
+			binary.LittleEndian.PutUint64(dst[offset:], k^v)
+		}
+	}
+
+	for i := (n - n%wordSize); i < n; i++ {
 		dst[i] = a[i] ^ b[i]
 	}
 }