You have a mixed-integer linear programming problem.
You’re tired of free solvers being like 100x slower than commercial solvers, right?
I think I found the culprit. It’s hard to contribute to free solvers despite them being easy to test. So, let’s add a small catalyst to the change you want to happen. I will try to turn it into a game that people can play. No fancy review or anything, just accepting objective performance gain.
If you want to kickstart the change, you can do so. Welcome to the craze.
Note: I am not an expert in MIP solvers. My only job is to accept pull requests that can kickstart the project in the first stage or improve the performance of the solver once that’s done. I only seek to be a small catalyst for anyone wanting to make such a thing happen. In all likelihood, this project goes nowhere, and it does not claim otherwise. It probably takes more effort than just starting a blank repo to kickstart the project, and even if it goes somewhere, in all likelihood, it would fail to compete. The world is not easy. Treat this as a lottery ticket and no more. On the off chance that someone decides to take what’s practically a joke and perhaps take the joke seriously, perhaps as a joke in itself, however, I would conclude we are in a funny timeline.
Probably so. I’m not sure if it would work either because where I came from (Chess), Stockfish beat commercial chess engines by a huge margin by allowing open contribution, where anyone is free to contribute an idea. However, chess is a fun state space search problem where one can easily have and test their ideas, given enough computing power. I’m not sure how hard it would be to do so on a solver. But as I said before, it probably would not work.
But well, miracles happen. I’m just a bit annoyed that free solvers are not just slightly, but like 100x slower than commercial solvers. That’s painful. Unfortunately, you can’t easily prune branches with heuristics like in Chess and so on. You want an optimal solution. I understand that optimization is unforgiving. It was all fun and games for me until it wasn’t anymore. I thought a computer search would at least outperform me in some silly problem, but I still outperformed it in my problem (reformulated as an MILP for my solver).
Take a look at the code. It might be slower than commercial solvers but it is not a trivial project where it is easy to make contributions or inprovements!
The fact is that I do not even know what is the real bottleneck in HiGHS development. Adding more devs or making it easier to develop so more devs would join would only help if dev count is the bottleneck and so on. If the bottleneck is the test set or the compute power needed to run tests, for example, then you need a different plan.