• 1 reply
  • Latest Post - ‏2013-08-02T17:54:00Z by AlexFleischer
25 Posts

Pinned topic OPL vs Concert vs Callable Library performance

‏2013-07-31T11:47:28Z |


I am wondering how much could I gain by implementing my OPL model with Concert or Callable library.

I am dealing with a problem where I am solving a succession of fairly simple MILP problems (hundreds to low tens of thousands rows and collumns). The MILP problems are well formulated so they get solved very quickly, the problem is that I have quite a few hundred of these problems to solve, so saving even 0.1 second per problem would be interesting for me.

I imagine that the price that I am paying for all the nice features in OPL is some computation overhead that is necessary to convert the model and send it to CPLEX in the right format. I would like to know if you have any idea what this overhead is : lets say I am solving my OPL model from the command line and it takes me a second to get the optimal solution. Is there a ball park figure how much could I save with Concert or callable library? 10%? 50%?

Obviously I know that the only way to really know is to implement my model in both frameworks but that  would take me a couple of weeks. As I have other life interests for my weekends than developping in C, I would like to give it some thought first...

I would be gratefull for any pointers..


  • AlexFleischer
    3151 Posts

    Re: OPL vs Concert vs Callable Library performance




    as we discussed over the phone, there are time and memory overheads with the OPL and concert layer but

    Solve time is huger than extraction time anyway most of the time
    Memory is cheap

    Plus we (human beings) tend to have better very good ideas when we look at a high level language model than when we look at a low level matrix description.