Re: [Help-glpk] glpsol, arbitrary precision and large numbers

[Andrew Makhorin]

> I see. One further question; am I likely to run into similar problems
> using other solvers?

Undoubtely, until the solver uses exact arithmetic. (It does not
necessarily mean using rational numbers; for example, the minisat solver
available with the --minisat option converts 0-1 mip to a satisfiability
problem which is then solved with a specialized search algorithm where
the lp relaxation is not used and therefore round-off errors do not
appear.)

> 
> For example PPL has been used to verify C and C++ code.  Thus, 32 and
> 64 bit numbers mist be analysed. 
> 

AFAIK, ppl uses bignums (arbitrary precision arithmetic), not
floating-point arithmetic.