The choice of optimal algorithms for fast computation of functions are of great interest in numerical theory and have increasing importance in the organisation and development of computer systems. The main problem here is the definition of such computational methods of universal character in use and such functional property that justifies the hardware implementation. This study examines the process of solving mathematical functions on computers using Horner's scheme and continued fraction. It concludes that procedures combining both algorithms may be efficiently used as an option in software for the approximation of linear systems of equations, serial and pipelined functions.