Compare what matters.

Choose up to four models. Test results are grouped by protocol version and physical or simulated execution. Read the conditions before treating results as comparable.