Hello;

For: nvar=60, x1=0.02, x2=0.03, eps=1.e-6, h1=0.009, hmin=0.001
and single precision, I'm having tremendous difficulty in getting any results!!

Here're the basic observations:
a) no "stepsize underflow" custom messages from RKQC
b) successive "next step" hnext is in the order of 10^(-8)
c) hnext is not generally increasing or decreasing, e.g.;
...........
4.33E-8
3.10E-8
2.86E-8
2.61E-8
2.64E-8
2.76E-8
2.98E-8
3.84E-8
3.60E-8
3.81E-8
4.35E-8
...........
4.75E-8
5.31E-8
5.15E-8
5.35E-8
3.96E-8
4.06E-8
3.67E-8
4.61E-8
...........
1.78E-8
2.20E-8
2.14E-8
2.00E-8
1.75E-8
1.63E-8
1.54E-8
1.56E-8
1.58E-8
1.50E-8
1.76E-8
2.12E-8
1.93E-8
...........

d) no "normal exit" from ODEINT after 10s of hrs on 3.16 GHz m/c, and nstp in the 1,000s

Is it possible to successfully exit ODEINT when ALL elements of the dependent variable vector remain unchanged from step to step (say, |diff| < eps)??? even when the stepsize is not producing underflow or is smaller than the input "guessed" hmin.

If it is possible, then How to do so without disrupting the dataflow ???

It is unfortunate that this Forum (F90) is no longer sponsored by the NR Authors, and the fortran code is not available for NR3 with its new (and possibly enhanced/corrected) algorithms.

Thank you kindly for your expert suggestion.

Monir

Supplementary:

I forgot to mention in my thread that the solution of the same aerodynamic problem is achievable with ~ 80% satisfaction when applying the classic RK 4th order integration scheme.
It's my hope to improve on the results by 'successfully' applying the NR adaptive stepsize control method.

Your advice/help would be greatly appreciated!

Monir