diff --git a/src/port.h b/src/port.h
index bac47a9..b302d6c 100644
--- a/src/port.h
+++ b/src/port.h
@@ -60,4 +60,15 @@ inline int Log2Floor(uint32 n) {
 #define PREDICT_TRUE(x) (x)
 #endif
 
+#if (defined(__ARM_ARCH) && (__ARM_ARCH == 7)) || \
+    (defined(M_ARM) && (M_ARM == 7)) || \
+    defined(__aarch64__) || defined(__ARM64_ARCH_8__) || defined(__i386) || \
+    defined(_M_IX86) || defined(__x86_64__) || defined(_M_X64)
+#if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
+#define WOFF_LITTLE_ENDIAN
+#elif defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
+#define WOFF_BIG_ENDIAN
+#endif  /* endianness */
+#endif  /* CPU whitelist */
+
 #endif  // WOFF2_PORT_H_
diff --git a/src/store_bytes.h b/src/store_bytes.h
index 74b279c..82c1bd8 100644
--- a/src/store_bytes.h
+++ b/src/store_bytes.h
@@ -22,6 +22,8 @@
 #include <stddef.h>
 #include <string.h>
 
+#include "./port.h"
+
 namespace woff2 {
 
 inline size_t StoreU32(uint8_t* dst, size_t offset, uint32_t x) {
@@ -33,10 +35,10 @@ inline size_t StoreU32(uint8_t* dst, size_t offset, uint32_t x) {
 }
 
 inline size_t Store16(uint8_t* dst, size_t offset, int x) {
-#if (defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__))
+#if defined(WOFF_LITTLE_ENDIAN)
   *reinterpret_cast<uint16_t*>(dst + offset) =
       ((x & 0xFF) << 8) | ((x & 0xFF00) >> 8);
-#elif (defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__))
+#elif defined(WOFF_BIG_ENDIAN)
   *reinterpret_cast<uint16_t*>(dst + offset) = static_cast<uint16_t>(x);
 #else
   dst[offset] = x >> 8;
@@ -53,11 +55,11 @@ inline void StoreU32(uint32_t val, size_t* offset, uint8_t* dst) {
 }
 
 inline void Store16(int val, size_t* offset, uint8_t* dst) {
-#if (defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__))
+#if defined(WOFF_LITTLE_ENDIAN)
   *reinterpret_cast<uint16_t*>(dst + *offset) =
       ((val & 0xFF) << 8) | ((val & 0xFF00) >> 8);
   *offset += 2;
-#elif (defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__))
+#elif defined(WOFF_BIG_ENDIAN)
   *reinterpret_cast<uint16_t*>(dst + *offset) = static_cast<uint16_t>(val);
   *offset += 2;
 #else
diff --git a/src/woff2_common.cc b/src/woff2_common.cc
index eba6a46..c19acdb 100644
--- a/src/woff2_common.cc
+++ b/src/woff2_common.cc
@@ -18,6 +18,8 @@
 
 #include "./woff2_common.h"
 
+#include "./port.h"
+
 namespace woff2 {
 
 
@@ -25,11 +27,11 @@ uint32_t ComputeULongSum(const uint8_t* buf, size_t size) {
   uint32_t checksum = 0;
   size_t aligned_size = size & ~3;
   for (size_t i = 0; i < aligned_size; i += 4) {
-#if (defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__))
+#if defined(WOFF_LITTLE_ENDIAN)
     uint32_t v = *reinterpret_cast<const uint32_t*>(buf + i);
     checksum += (((v & 0xFF) << 24) | ((v & 0xFF00) << 8) |
       ((v & 0xFF0000) >> 8) | ((v & 0xFF000000) >> 24));
-#elif (defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__))
+#elif defined(WOFF_BIG_ENDIAN)
     checksum += *reinterpret_cast<const uint32_t*>(buf + i);
 #else
     checksum += (buf[i] << 24) | (buf[i + 1] << 16) |