Alexandre Amice

Per @jwnimmer-tri's suggestion for a function. > ```c++ > /** Solves prog[i] into result[i], optionally using initial_guess[i] and solver_options[i] if given. > Uses at most parallelism cores, with static scheduling...

Is specifying a `SolverId` strictly better than specifying a `SolverInterface`? The only thing I dislike in this is the difference in spelling from the vanilla `Solve` method. Traditionally `Solve(prog)` is...

I suppose the issue with: ``` /** Solves prog[i] into result[i], optionally using initial_guess[i] and solver_options[i] if given. If `solver_id` is given, then all programs will be solved using instances...

I have a PR up at #21957. I changed the function signature a bit more to avoid constructing too many solver objects.

cc @agarwal-abhinav who unearthed the bug in #21860 which was introduced in #21308

One more complicating factor is what should the return type be? Currently, `IrisInConfigurationSpaceFromCliqueCover` crashes if `IrisInConfigurationSpace` fails to return a region. The PR #21376 is stalled since we are **not**...

See #21879 for an alternative design based on classes

FWIW: [#247](https://github.com/coin-or/Clp/issues/247) claims there is an issue with the presolver when solving QPs in CLP and this program is trivial to pre-solve `x0` away. I haven't actually looked that this...

[Linking the discussion from the Drake Developers Slack on this subject](https://drakedevelopers.slack.com/archives/C0557S260N8/p1719572390231669)

I don't know if/when I'll have time to work this one out with tests and everything, but here's some code based on using `std::variant` to get you started if you...