The ARC cpu family currently does not have hardware floating point support. Software floating point support is provided by GCC and uses IEEE floating-point numbers.
GCC