Originally posted by: SystemAdmin
When you say different results, do you mean different optimal values, or one program solving to optimality and another not getting there? Or do you mean different optimal solutions with the same objective value (but different variable values)?
The sequence of pivots in solving an LP (and the sequence of nodes in solving a MIP) can be influenced by small differences in rounding, differences in the order that constraints and/or variables appear with in the model, phases of the moon, ... If a problem has multiple optimal solutions, OPL v. Java (or OPL script writing constraints C1 then C2 then C3 and Java program writing C1 then C3 then C2) or Win 64 v. Win 32 could easily cause differences in the final solution (and the path to it).
In particular, Win 64 may do double precision arithmetic slightly differently than Win 32, and OPL may do it slightly differently than Java. There are known differences in floating point across hardware platforms, between different flavors of C compilers, etc. They won't cause you to get a wrong answer, but the differences, subtle as they are, can alter the flow of the algorithms.
/Paul
Mathematicians are like Frenchmen: whenever you say something to them, they translate it into their own language, and at once it is something entirely different. (Goethe)
#CPLEXOptimizers#DecisionOptimization