The objective of this software is to provide a solver for Linear Programming (and Integer Programming to a lesser degree) that returns true (rational) optimal solutions.
It relies heavilly on the GNUMP library, that provides a multiprecision library for both floating point and also rational arithmetic. Note that if you use a dynamicly linked version of QSopt-Exact, then the GMP library should have been compiled with the option --enable-alloca=malloc-reentrant, this is needed to avoid memory corruption. The basis for the LP solver was taken from QSopt, which is an LP solver based on floating point arithmetic and available for free for research purposes. A brief description of the implementation and the obtained results can be obtained here (pdf), and a longer description (which is part of my Ph.D. thesis) can be found here Much of the functionality used in QSopt-Exact comes from EGlib.
You can see the documentation or download the program source.
We also have made available binaries for linux 32 bit and for linux 64 bit
Finally, many thanks to all people that have contributed with bug-reports, comments, and help.