I explain how to associate computations with physical systems such as finite degree of freedom hamiltonian systems, and thus show why we should consider such systems as real number machines. I make a few comparisons between these real number machines of those and Blum, Shub and Smale (the question of equivalence is still open), and also make some general comments on complexity in physics.
Publication information: My write-up of my talk was rejected for the proceedings. I unfortunately do not have the (single) referee report or editor's comments (they were lost in an email disaster some years back). I did not agree with their judgement at all, and would have appealed the decision, but the timeframe made such an appeal impossible.