Floating-point numbers

Is it possible to find different results for the same scenario running in different computers. Algorithms based on Mixed Integer Linear Programming (MILP) depend on third-party solvers and their results may differ in terms of floating-point precision from hardware to hardware.

Given that MILP is based on multiple LP executions, precision differences may accumulate over the sequence of operations performed with floating-point numbers. It’s expected that results may differ, but they should be equivalent in terms of NPV. If the physical results are too different, this means this mine has the flexibility to operate in both ways without big impacts in NPV.

For an in-depth explanation on floating-point arithmetic from a computer science perspective please see here.