diff --git a/src/util/compat/bswap.h b/src/util/compat/bswap.h
new file mode 100644
index 0000000..a8acdcf
--- /dev/null
+++ b/src/util/compat/bswap.h
@@ -0,0 +1,44 @@
+#pragma once
+
+#include <stdint.h>
+
+#if defined(__GNUC__) || defined(__clang__)
+#define bswap16 __builtin_bswap16
+#define bswap32 __builtin_bswap32
+#define bswap64 __builtin_bswap64
+#elif __has_include(<byteswap.h>)
+#include <byteswap.h>
+#define bswap16 bswap_16
+#define bswap32 bswap_32
+#define bswap64 bswap_64
+#else
+#define bswap16 bswap16_impl
+#define bswap32 bswap32_impl
+#define bswap64 bswap64_impl
+#define IMPL_BSWAP
+#endif
+
+#if defined(IMPL_BSWAP) || !defined(NDEBUG)
+static inline uint16_t bswap16_impl(uint16_t x) {
+	return (x << 8) | (x >> 8);
+}
+
+static inline uint32_t bswap32_impl(uint32_t x) {
+	return (x << 24) |
+		((0x0000FF00U & x) << 8) |
+		((0x00FF0000U & x) >> 8) |
+		(x >> 24);
+}
+
+static inline uint64_t bswap64_impl(uint64_t x) {
+	return (x << 56) |
+		((0x000000000000FF00ULL & x) << 40) |
+		((0x0000000000FF0000ULL & x) << 24) |
+		((0x00000000FF000000ULL & x) << 8) |
+		((0x000000FF00000000ULL & x) >> 8) |
+		((0x0000FF0000000000ULL & x) >> 24) |
+		((0x00FF000000000000ULL & x) >> 40) |
+		(x >> 56);
+}
+#endif
+#undef IMPL_BSWAP
diff --git a/src/util/compat/endian.h b/src/util/compat/endian.h
index f326815..a3f6d9e 100644
--- a/src/util/compat/endian.h
+++ b/src/util/compat/endian.h
@@ -5,18 +5,7 @@
 #include <stdint.h>
 
 #include "../atrb.h"
-
-#if __has_include(<byteswap.h>)
-#include <byteswap.h>
-#elif defined(__GNUC__) || defined(__clang__)
-#define bswap_16 __builtin_bswap16
-#define bswap_32 __builtin_bswap32
-#define bswap_64 __builtin_bswap64
-#else
-// TODO: use custom bswap bacros
-#error WIP, no support
-#endif
-
+#include "bswap.h"
 
 /* little endian */
 atrb_const static inline uint16_t le16ton(uint16_t); // converts little-endian (LE) encoding to native for a 16 bit integer. (NOOP if native is LE)
@@ -41,12 +30,12 @@ uint64_t le64ton(uint64_t x) { return x; }
 uint16_t ntole16(uint16_t x) { return x; }
 uint32_t ntole32(uint32_t x) { return x; }
 uint64_t ntole64(uint64_t x) { return x; }
-uint16_t be16ton(uint16_t x) { return bswap_16(x); }
-uint32_t be32ton(uint32_t x) { return bswap_32(x); }
-uint64_t be64ton(uint64_t x) { return bswap_64(x); }
-uint16_t ntobe16(uint16_t x) { return bswap_16(x); }
-uint32_t ntobe32(uint32_t x) { return bswap_32(x); }
-uint64_t ntobe64(uint64_t x) { return bswap_64(x); }
+uint16_t be16ton(uint16_t x) { return bswap16(x); }
+uint32_t be32ton(uint32_t x) { return bswap32(x); }
+uint64_t be64ton(uint64_t x) { return bswap64(x); }
+uint16_t ntobe16(uint16_t x) { return bswap16(x); }
+uint32_t ntobe32(uint32_t x) { return bswap32(x); }
+uint64_t ntobe64(uint64_t x) { return bswap64(x); }
 #elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
 uint16_t le16ton(uint16_t x) { bswap_16(x); }
 uint32_t le32ton(uint32_t x) { bswap_32(x); }