Small Grant Request+Completion: DAE Problem Benchmarks

[jayantpranjal0]

Reference PR: SciML/SciMLBenchmarks.jl#1303

[ChrisRackauckas]

This one isn't merged yet.

[ChrisRackauckas]

Merged now 🎉 so this is done

[jayantpranjal0]

Thanks😊

[ChrisRackauckas]

What is the new request here?

[jayantpranjal0]

I was thinking of writing another dae benchmark(Wheelset problem).

[ChrisRackauckas]

Okay sounds good to me. Note that because it's an Index-2 problem, you'll likely need to use ModelingToolkit to derive the Index-1 problem for the mass matrix and DAE problem form, and then also have a separate implementation of the index-2 problem form in direct code, and then work out which solvers end up being able work if the correct index-2 initialization is given. So that will have a few more complications from before, but it's not really well-characterized how the methods will do here (Rodas5P and radau should be able to handle index-2 IIRC, https://www.sciencedirect.com/science/article/abs/pii/S0168927496000232), so it'll be very useful to have a benchmark which finally stresses this behavior and gives a good categorization of index-2 handling, along with benchmarks to see if there is performance lost by ModelingToolkit's reduction to index-1.

Definitely approved here under the understanding the benchmark will include this as a required element in the benchmarking.

[ChrisRackauckas]

Requires two other approvals.