Yes after updating to kernel 1.2.0, I am seeing much better results. Thanks for all the replies.
If anyone is interested, here are the new results for a bunch of solvers (compared to FLINT). I ran the script 3 times and these are the numbers from the third run on my computer. Of course, these results apply to this particular set of differential equations and initial conditions.
tol: 1.0e-11
Computational time in seconds
- DOPRI5: 1.38711e+01 (vs 2.19 Fortran)
- DP5: 6.26217e+00 (-do-)
- TSIT5: 7.09554e+00 (-do-)
- DOP853: 3.98887e+00 (vs 0.73 Fortran)
- DP8: 3.54820e+00 (-do-)
- Vern6: 5.17847e+00 (vs 2.0 in Fortran)
- Vern9: 5.01823e+00 (vs 0.98 in Fortran)