Once the dependency upgrade in #826 (tracked from #815) is merged, we should build out a proper problem bank: a collection of optimal control problems solved through both the direct and indirect methods, on both CPU and GPU, run on a periodic basis (e.g. nightly or weekly CI) rather than only as part of the regular test suite.
The upgrade PR only wires up a handful of problems in dual (:abstract, :functional) form to unblock the migration — this issue is about growing that into a broader, recurring regression bank (many more problems, reference solutions, CPU/GPU parity checks) once things have settled.
No need to spec out the details here — this is a placeholder to come back to.
Once the dependency upgrade in #826 (tracked from #815) is merged, we should build out a proper problem bank: a collection of optimal control problems solved through both the direct and indirect methods, on both CPU and GPU, run on a periodic basis (e.g. nightly or weekly CI) rather than only as part of the regular test suite.
The upgrade PR only wires up a handful of problems in dual
(:abstract, :functional)form to unblock the migration — this issue is about growing that into a broader, recurring regression bank (many more problems, reference solutions, CPU/GPU parity checks) once things have settled.No need to spec out the details here — this is a placeholder to come back to.