diff --git a/ConfigureChecks.cmake b/ConfigureChecks.cmake
index 5f4f6d9d..0eb04018 100644
--- a/ConfigureChecks.cmake
+++ b/ConfigureChecks.cmake
@@ -1,8 +1,6 @@
 include(CheckLibraryExists)
 include(CheckTypeSize)
 include(CheckCXXSourceCompiles)
-include(TestBigEndian)
-include(TestFloatFormat)
 
 # Check if the size of numeric types are suitable.
 
@@ -36,27 +34,6 @@ if(NOT ${SIZEOF_DOUBLE} EQUAL 8)
   message(FATAL_ERROR "TagLib requires that double is 64-bit wide.")
 endif()
 
-# Determine the CPU byte order.
-
-test_big_endian(IS_BIG_ENDIAN)
-
-if(NOT IS_BIG_ENDIAN)
-  set(SYSTEM_BYTEORDER 1)
-else()
-  set(SYSTEM_BYTEORDER 2)
-endif()
-
-# Check if the format of floating point types are suitable.
-
-test_float_format(FP_IEEE754)
-if(${FP_IEEE754} EQUAL 1)
-  set(FLOAT_BYTEORDER 1)
-elseif(${FP_IEEE754} EQUAL 2)
-  set(FLOAT_BYTEORDER 2)
-else()
-  message(FATAL_ERROR "TagLib requires that floating point types are IEEE754 compliant.")
-endif()
-
 # Determine which kind of atomic operations your compiler supports.
 
 check_cxx_source_compiles("
diff --git a/cmake/TestFloatFormat.c b/cmake/TestFloatFormat.c
deleted file mode 100644
index 4a7b32e4..00000000
--- a/cmake/TestFloatFormat.c
+++ /dev/null
@@ -1,17 +0,0 @@
-int main(int argc, char **argv)
-{
-  int ret = 0;
-
-  double bin1[] = {
-    // "*TAGLIB*" encoded as a little-endian floating-point number
-    (double)3.9865557444897601e-105, (double)0.0
-  };
-  float bin2[] = {
-    // "*TL*" encoded as a little-endian floating-point number
-    (float)1.81480400e-013, (float)0.0
-  };
-  ret += ((int*)bin1)[argc];
-  ret += ((int*)bin2)[argc];
-
-  return ret;
-}
diff --git a/cmake/modules/TestFloatFormat.cmake b/cmake/modules/TestFloatFormat.cmake
deleted file mode 100644
index ef9b1860..00000000
--- a/cmake/modules/TestFloatFormat.cmake
+++ /dev/null
@@ -1,61 +0,0 @@
-# Returns 1 if IEEE754 little-endian, 2 if IEEE754 big-endian, otherwise 0.
-
-MACRO(TEST_FLOAT_FORMAT FP_IEEE754)
-  IF(NOT FP_IEEE754)
-    TRY_COMPILE(HAVE_FLOAT_FORMAT_BIN
-      "${CMAKE_BINARY_DIR}" "${CMAKE_CURRENT_SOURCE_DIR}/cmake/TestFloatFormat.c"
-    COPY_FILE "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/TestFloatFormat.bin")
-
-    SET(FP_IEEE754 0)
-
-    IF(HAVE_FLOAT_FORMAT_BIN)
-
-      # dont match first/last letter because of string rounding errors :-)
-      FILE(STRINGS "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/TestFloatFormat.bin"
-        DOUBLE_IEEE754_LE LIMIT_COUNT 1 REGEX "TAGLIB")
-      FILE(STRINGS "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/TestFloatFormat.bin"
-        DOUBLE_IEEE754_BE LIMIT_COUNT 1 REGEX "BILGAT")
-      FILE(STRINGS "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/TestFloatFormat.bin"
-        FLOAT_IEEE754_LE LIMIT_COUNT 1 REGEX "TL")
-      FILE(STRINGS "${CMAKE_BINARY_DIR}${CMAKE_FILES_DIRECTORY}/TestFloatFormat.bin"
-        FLOAT_IEEE754_BE LIMIT_COUNT 1 REGEX "LT")
-
-      IF(DOUBLE_IEEE754_LE AND FLOAT_IEEE754_LE)
-        SET(FP_IEEE754_LE 1)
-      ENDIF()
-
-      IF(DOUBLE_IEEE754_BE AND FLOAT_IEEE754_BE)
-        SET(FP_IEEE754_BE 1)
-      ENDIF()
-
-      # OS X Universal binaries will contain both strings, set it to the host
-      IF(FP_IEEE754_LE AND FP_IEEE754_BE)
-        IF(CMAKE_SYSTEM_PROCESSOR MATCHES powerpc)
-            SET(FP_IEEE754_LE FALSE)
-            SET(FP_IEEE754_BE TRUE)
-        ELSE()
-            SET(FP_IEEE754_LE TRUE)
-            SET(FP_IEEE754_BE FALSE)
-        ENDIF()
-      ENDIF()
-
-      IF(FP_IEEE754_LE)
-        SET(FP_IEEE754 1)
-      ELSEIF(FP_IEEE754_BE)
-        SET(FP_IEEE754 2)
-      ENDIF()
-    ENDIF()
-
-    # just some informational output for the user
-    IF(FP_IEEE754_LE)
-       MESSAGE(STATUS "Checking the floating point format - IEEE754 (LittleEndian)")
-    ELSEIF(FP_IEEE754_BE)
-       MESSAGE(STATUS "Checking the floating point format - IEEE754 (BigEndian)")
-    ELSE()
-       MESSAGE(STATUS "Checking the floating point format - Not IEEE754 or failed to detect.")
-    ENDIF()
-
-    SET(FP_IEEE754 "${${FP_IEEE754}}" CACHE INTERNAL "Result of TEST_FLOAT_FORMAT" FORCE)
-  ENDIF()
-ENDMACRO(TEST_FLOAT_FORMAT FP_IEEE754)
-
diff --git a/config.h.cmake b/config.h.cmake
index 44cabfce..1bec58cc 100644
--- a/config.h.cmake
+++ b/config.h.cmake
@@ -1,13 +1,5 @@
 /* config.h. Generated by cmake from config.h.cmake */
 
-/* Integer byte order of your target system */
-/* 1 if little-endian, 2 if big-endian. */
-#cmakedefine   SYSTEM_BYTEORDER ${SYSTEM_BYTEORDER}
-
-/* IEEE754 byte order of your target system. */
-/* 1 if little-endian, 2 if big-endian. */
-#cmakedefine   FLOAT_BYTEORDER  ${FLOAT_BYTEORDER}
-
 /* Defined if your compiler supports some byte swap functions */
 #cmakedefine   HAVE_GCC_BYTESWAP_16 1
 #cmakedefine   HAVE_GCC_BYTESWAP_32 1
diff --git a/taglib/toolkit/tbytevector.cpp b/taglib/toolkit/tbytevector.cpp
index 45ad0317..6262bec6 100644
--- a/taglib/toolkit/tbytevector.cpp
+++ b/taglib/toolkit/tbytevector.cpp
@@ -209,7 +209,7 @@ T toNumber(const ByteVector &v, size_t offset, size_t length, bool mostSignifica
 template <class T>
 T toNumber(const ByteVector &v, size_t offset, bool mostSignificantByteFirst)
 {
-  static const bool isBigEndian = (Utils::SystemByteOrder == Utils::BigEndian);
+  const bool isBigEndian = (Utils::systemByteOrder() == Utils::BigEndian);
   const bool swap = (mostSignificantByteFirst != isBigEndian);
 
   if(offset + sizeof(T) > v.size())
@@ -228,7 +228,7 @@ T toNumber(const ByteVector &v, size_t offset, bool mostSignificantByteFirst)
 template <class T>
 ByteVector fromNumber(T value, bool mostSignificantByteFirst)
 {
-  static const bool isBigEndian = (Utils::SystemByteOrder == Utils::BigEndian);
+  const bool isBigEndian = (Utils::systemByteOrder() == Utils::BigEndian);
   const bool swap = (mostSignificantByteFirst != isBigEndian);
 
   if(swap)
@@ -251,7 +251,7 @@ TFloat toFloat(const ByteVector &v, size_t offset)
   } tmp;
   ::memcpy(&tmp, v.data() + offset, sizeof(TInt));
 
-  if(ENDIAN != Utils::FloatByteOrder)
+  if(ENDIAN != Utils::floatByteOrder())
     tmp.i = Utils::byteSwap(tmp.i);
 
   return tmp.f;
@@ -266,7 +266,7 @@ ByteVector fromFloat(TFloat value)
   } tmp;
   tmp.f = value;
 
-  if(ENDIAN != Utils::FloatByteOrder)
+  if(ENDIAN != Utils::floatByteOrder())
     tmp.i = Utils::byteSwap(tmp.i);
 
   return ByteVector(reinterpret_cast<char *>(&tmp), sizeof(TInt));
diff --git a/taglib/toolkit/tstring.cpp b/taglib/toolkit/tstring.cpp
index 15db518d..258e1fcf 100644
--- a/taglib/toolkit/tstring.cpp
+++ b/taglib/toolkit/tstring.cpp
@@ -828,7 +828,7 @@ void String::copyFromUTF16(const char *s, size_t length, Type t)
 }
 
 const String::Type String::WCharByteOrder
-  = (Utils::SystemByteOrder == Utils::BigEndian) ? String::UTF16BE : String::UTF16LE;
+  = (Utils::systemByteOrder() == Utils::BigEndian) ? String::UTF16BE : String::UTF16LE;
 
 }
 
diff --git a/taglib/toolkit/tutils.h b/taglib/toolkit/tutils.h
index c5e0cc77..68e5fc56 100644
--- a/taglib/toolkit/tutils.h
+++ b/taglib/toolkit/tutils.h
@@ -216,20 +216,9 @@ namespace TagLib
       BigEndian
     };
 
-#ifdef SYSTEM_BYTEORDER
-
-# if SYSTEM_BYTEORDER == 1
-
-    const ByteOrder SystemByteOrder = LittleEndian;
-
-# else
-
-    const ByteOrder SystemByteOrder = BigEndian;
-
-# endif
-
-#else
-
+    /*!
+     * Returns the integer byte order of the system.
+     */
     inline ByteOrder systemByteOrder()
     {
       union {
@@ -244,41 +233,26 @@ namespace TagLib
         return BigEndian;
     }
 
-    const ByteOrder SystemByteOrder = systemByteOrder();
-
-#endif
-
-#ifdef FLOAT_BYTEORDER
-
-# if FLOAT_BYTEORDER == 1
-
-    const ByteOrder FloatByteOrder = LittleEndian;
-
-# else
-
-    const ByteOrder FloatByteOrder = BigEndian;
-
-# endif
-
-#else
-
+    /*!
+     * Returns the IEEE754 byte order of the system.
+     */
     inline ByteOrder floatByteOrder()
     {
-        double bin[] = {
-            // "*TAGLIB*" encoded as a little-endian floating-point number
-            (double) 3.9865557444897601e-105, (double) 0.0
-        };
+      union {
+        double d;
+        char   c;
+      } u;
 
-        char *str = (char*)&bin[0];
-        if(strncmp(&str[1], "TAGLIB", 6) == 0)
-          return LittleEndian;
-        else
-          return BigEndian;
+      // 1.0 is stored in memory like 0x3FF0000000000000 in canonical form.
+      // So the first byte is zero if little endian.
+
+      u.d = 1.0;
+      if(u.c == 0)
+        return LittleEndian;
+      else
+        return BigEndian;
     }
 
-    const ByteOrder FloatByteOrder = floatByteOrder();
-
-#endif
   }
 }