Benchmarking

[ocots]

This discussion is the main page about benchmarking, which is one of the main goal of the roadmap 2025-26.

In progress: MAM4 Project.

To setup dev config and contribute, visit our Wiki.

We welcome @vmerc!

[ocots]

For the benchmarking, we have:

• OptimalControlProblems.jl: A collection of optimal control problems with ODE's in Julia.
• CTBenchmarks.jl: the repo where the benchmarks are computed and presented.

[ocots]

The different benchmarks we intend to do are:

• Comparison of combos (modeler, solver) on CPU and GPU and CRU 🤣
• Comparison of combos (scheme, solver) for the best modeler on CPU and GPU
• ...

Play on AD backend, options.

[jbcaillau]

welcome @vmerc !

[jbcaillau]

@ocots @PierreMartinon a nice trick (thanks @amontoison) to connect via ssh / tmate to an ongoing runner (the workflow prints out the ssh command to use):

https://github.com/mxschmitt/action-tmate

[jbcaillau]

cc @rveltz @j-l-s