As a test for something, I am trying to see if I can implement column generation (true to my nature / background 😉 ) in AIMMS. The pricing problem I am looking at is some sort of a knapsack problem with some extra constraints.
Of course this can easily be modeled as an ILP that I solve, but the overhead for this seems to be pretty large. Another problem is that I potentially need to solve quite some pricing problems in every iteration of the column generation process.
Â
The pricing problem lends itself perfectly to be implemented in some basic python, so I looked at the aimms python bridge and that also works. However, I now wanted to look at the threading-free implementation in mostly python 3.14 that would allow me to do actual multi-threading within the same process in python. Since each individual pricing problem that needs to be solved during one iteration is independent , it could be solved in a separate thread. This would allow me to scale this perfectly to the number of cores available.
Â
However, this requires me to have access to a distributed version of the aimmspy library that is built for the free-threading (having something like cp14t in the name). Is there any view on whether this is going to be made available?