i sometimes wrap existing C interfaces to make solvers available from the JuMP modeling language in Julia.
This would allow easy use of your solver by people and other solvers in the JuMP ecosystem.
This would require me to:
Find a reproducible (cross platform) way to build binaries
Call the C interface from Julia (while calling Python from Julia is also possible is expect no benefits from using that interface)
Hook up the Julia version of your C Interface to the algebraic modeling language
Maintain this code base
Since this is would require work on my behalf, i would like to ask some questions:
Are you even interested in your solver being made readily available?
Would you in general recommend reuse of your code or would a clean reimplemtation be a better way?
Do you suspect your solver to has competitive performance on non mixed integer instances?
Do you have some benchmarks comparing your solver to similar solvers (your code has a toggle for Gurobi?)?
A QP solver allows solving NLP via SQP. Can a simple code on top of your solver be used in a similar manner to solve MINLP problems or does that sound very inefficient?
The practice seen in input.h to compile the solver for each problem, while sensible for embedded work, doesn't really work for user specified problems. You being more familiar with your code base, is there a way to specify problems at run time and if not do you suspect it being worth the making a nicer C interface?
You algorithm depends on an parameter i would have to expose to the user, unless there are good automatic heuristic to choose it. Do you have any insights on that?
Hello,
i sometimes wrap existing C interfaces to make solvers available from the JuMP modeling language in Julia. This would allow easy use of your solver by people and other solvers in the JuMP ecosystem. This would require me to:
Since this is would require work on my behalf, i would like to ask some questions: