David Gardner

@pazner Thanks for the ping, I meant to take a look at this last week. I've got some time this afternoon to review.

> @cjvogl I'm a bit confused on the implicit solve. RK stages in the more common context of `u_{n+1} = u_n + dt*\sum b_ik_i` require solving systems `k = f(x+dt*k,...

The CI did not catch this because there's a typo in `test/config_cmake.py`. `ENABLE_SUPERLU_DIST` should be `ENABLE_SUPERLUDIST`.

The number of function evaluations will depend on several factors (e.g., tolerances, nonlinear/linear solver, etc.) and dynamics of the problem your are solving. What nonlinear/linear solver are you using? If...

Currently the Picard iteration does not utilize the same linear system reuse controls that are present when using the modified Newton iteration. As such, the Picard iteration evaluates the linear...

Yes, let's keep this issue open.

Yes, that's correct. A user-supplied Jacobian function is currently required when using the oneMKL interface. This function may compute the analytic Jacobian or an approximation to it e.g., using difference-quotients.

> However, I still do not understand where to define the integrand rhs when rhsQS =NULL. Maybe CVodeQuadInit has to be called, but it is mentionned nowhere in the doc....

Since `SUNDIALSConfig.cmake` lists these options as off I'm guessing SUNDIALS was configured with the default settings. In this case any targets related to third-party libraries (i.e., KLU, hypre, MAGMA, and...

@balos1 I agree it would be good to to have separate the macros for the F77 interfaces and LAPACK. It does happen (though I've not run into it before) that...