Nowadays, most floating point computations are performed in double precision – with a mantissa of 53 bits. Nevertheless, depending on the application we are using, it is preferable more precision, such as double extended – 64 bits – quadruple extended – 113 bits – or even more. The main goal of this project is to research how we can achieve such arbitrary precision on PRISM – by this, we mean the user should be able to choose the precision of each calculation – in order to provide a close approximation of exact solutions.

Organization

Student

Nicolas Del Piano

Mentors

  • Vojtech Forejt
close

2016