WOG Symposium GENT 2004
- Official site : http://www.wogsymposium.ugent.be/
Multiprecision Matlab Environment
It is well-known that the Matlab software environment, while extremely powerful and of the highest standard, supports numerical computations only in the standard hardware precisions: floating-point single (32 bits), double (64 bits), integers (up to 64 bits) and complex doubles. For specific applications requiring higher precision, Matlab and all the numerical software it offers cannot be used. In this talk we present an interpreter and compiler which implements a large subset of the Matlab language. The advantage of our interpreter is that all computations can be performed in the fully IEEE compliant multiprecision, multibase floating-point library MpIeee, developed as part of the Arithmos project.
Besides the multiprecision floating-point datatype MpIeee, also rational, big integer (limited only by memory) and complex multiprecision datatypes are supported. Future extensions include multiprecision interval arithmetic. Inspired by the toolbox philosophy of Matlab, we have developed a plugin system, called LibLoader. Using this plugin system, a multiprecision software library of numerical routines can be loaded into our interpreter. We will demonstrate the above functionality by plugging in the multiprecision version of Numerical Recipes as well as a multiprecision version of the Gnu Scientific Library. Both these multiprecision versions are generated automatically from the original libraries using a precompiler which we have also developed.
- Slides : wog-Schreppers.pdf