Originally posted by: SystemAdmin

Hi to everyone,

I have a few questions related to adding cuts to a MIP model in CPLEX 12.1 using C++ API. I'm trying to add a quite big cut (approx. 35000 variables in the cut, 900000 variables in the model) at the root node and I've identified that it takes a very long time (+2 minutes) when doing add(cut).

The cut is created following the User's Manual recomendations (using an IloRange) and it is the first cut to be added after the first relaxation solved. The strange thing is that the overall time required for constructing the initial model is less than 2 minutes. Even creating the cut requires less than a second. Just in case, setting the Normalizer to IloFalse does not produces significant changes.

Is there any explanation for this behaviour? Would it be faster if using the Callable Library instead of C++ API?

Thanks in advance for any help or thought in this regard,

Juan.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Does it help to create the expression for the cut using IloScalProd()?
Are you adding the cut in a callback or after solve() returned?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Dear Daniel,

Thank you very much for your reply. I was not aware of this function. I'll try your suggestion to see if times are reduced. Regarding your question, I'm adding the cut in a callback.

Thanks again for your reply,

Juan.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

2 minutes seem very long for adding a single cut. Could you post your model and the cut here or send it to me daniel(dot)junglas(at)de(dot)ibm(dot)com.
For the model the best would be a SAV file. For the cut maybe just send what is output when print the cut on cout.
I could then use a profiling enabled version of CPLEX to see where the time is burnt.
#CPLEXOptimizers
#DecisionOptimization