[PekkaRo]

Dear all,

I am wondering what would be the best practice to declare fields that I will need to reuse, but which are not part of the output?

What I wonder is that in the parts, that are likely well optimized, some of the fields seem to be reallocated on each function call. For example, in the kEpsilon-model's correct() function, generation term G seems to be allocated at each time step. Most other fields are data members allocated at the initializer list. The latter seems better programming practice, but maybe there is a hidden catch I don't see?

I have inherited some models where the temporary allocation is used heavily, and I suspect that this might cause performance damage. Would this be the case? I am especially interested in large models, say >10^9 cells.