Enhancement of performance and accuracy of computation using rational type
Abdulwahab Mohammed Otay, MOIEZ A. TAPIA · 1996
Both the extent of use and the size of computer programs used in research, industry, business, and control process have been increasing exponentially. This necessitates enhancement of accuracy and speed of computation. Most commonly used type of numbers used in computation is floating-point which is inherently inexact in most cases. Theoretical work done on rational number systems has established that computation done using rational format can be more accurate than floating-point format in many cases. The work done in this thesis improves upon the software modules developed for computation done earlier using rational type and presents new parallel software modules in $\varepsilon$C, a dialect of C, that enhance the speed of computation.