jmyrberg
Check problem size limitations for all algorithms
Ensure correct errors are thrown regarding problem size limitations before trying to solve it within Fortran
It would be nice if it were possible to increase the limits slightly, at least for the "mtm" method for multiple knapsack. Finding a solution with 10 knapsacks and 85 items and profits of 1 took ~1.1 seconds to solve 1000 times (I placed it in a loop to better eliminate noise). Later on, I will likely need ~15-20 knapsacks, and it seems impossible to change the limit without forking and changing both the Fortran and Python.
Would it be reasonable to allow users to use up to 20-25 knapsacks, but display a warning to stdout (on by default but disabled if the user passes warnings=False) in case the user needs more than 10 knapsacks?
It would be nice if it were possible to increase the limits slightly, at least for the "mtm" method for multiple knapsack. Finding a solution with 10 knapsacks and 85 items and profits of 1 took ~1.1 seconds to solve 1000 times (I placed it in a loop to better eliminate noise). Later on, I will likely need ~15-20 knapsacks, and it seems impossible to change the limit without forking and changing both the Fortran and Python.
Would it be reasonable to allow users to use up to 20-25 knapsacks, but display a warning to stdout (on by default but disabled if the user passes
warnings=False) in case the user needs more than 10 knapsacks?
I have no objections in increasing the limits. However, if I remember correctly, I actually tried it, but it wasn’t as easy as I thought on the Fortran side, and thus left the code as-is. If you’re willing to have a go on this, feel free to fork and make a PR 🙂