Well, there is one simple brute-force way to see if you are getting either exactly the same thing (which is problematical, given that round-off error is a reality with computer arithmetic), or within tolerances for round-off error: use I/O API M3Tools program m3diff:
- if the per-timestep root mean square simple difference (“A-B”) is zero, then the data are identical;
- if the per-timestep root mean square normalized difference ("(A-B)/((A+B)/2)") is less than 1.0e-6, then the files agree to within round-off error tolerances