Some possibilities for the reduction of rounding errors in the integration of ordinary differential equations

The possibility of compensating the basic rounding errors affecting the accuracy of the numerical integration of ordinary differential equations, without the use of double digitry, is discussed.