diff --git a/library/common.h b/library/common.h
index 9c09602..94b8c5d 100644
--- a/library/common.h
+++ b/library/common.h
@@ -131,10 +131,10 @@
 #if defined(MBEDTLS_EFFICIENT_UNALIGNED_ACCESS)
 #if defined(__aarch64__) && defined(__ARM_NEON)
     for (; (i + 16) <= n; i += 16) {
-        uint64x2_t v1 = vld1q_u64((uint64_t *) a);
-        uint64x2_t v2 = vld1q_u64((uint64_t *) b);
+        uint64x2_t v1 = vld1q_u64((uint64_t *) (a + i));
+        uint64x2_t v2 = vld1q_u64((uint64_t *) (b + i));
         uint64x2_t x = veorq_u64(v1, v2);
-        vst1q_u64((uint64_t *) r, x);
+        vst1q_u64((uint64_t *) (r + i), x);
     }
 #elif defined(__amd64__) || defined(__x86_64__) || defined(__aarch64__)
     /* This codepath probably only makes sense on architectures with 64-bit registers */