Quadruple-precision floating-point format