The GMD standard is "scientific reproducibility", which means that at least the scientific results of a paper must be reproducible, even if bitwise reproducibility cannot not be assured. ie. computational differences due to different computational frameworks need to be small in comparison to the main findings.

In this case, the main result of the work is that one computational software system is faster than another (MPI beats OPM). It remains possible, as raised by one of the reviewers, that this may be due to computer architecture or other factors external the actual software system itself. I do not think this has been addressed in your revised manuscript, and would like to see a few sentences or a paragraph (or small section) devoted to discussion of this issue. For example, have you run the same experiments on a variety of architectures? If not, is it possible that another architecture could obtain the opposite result? If this matter can be trivially dispensed with, then just a short explanation in the discussion is all that would be required. Otherwise it is surely important to raise this as a caveat in the conclusions as well as discuss it in the main body of the paper.

Thanks for clarifying the computer architecture. I am now happy to accept the paper.