Double-precision floating-point format