Originally posted by: clemsonmacrone

Hi, I am having a memory leak at the makeBranch() lines of my branch callback routine. Is their something fundamentally wrong with my set up that causes this? note: F2Arcs is a vector of the variables I am branching on, and vars_ is a pointer to the variables in the model
IloExpr e1(getEnv());

for(int i =0; i < F2Arcs.size(); i++)
{
e1 += (*vars_)[F2Arcsi];
}

makeBranch( e1 == 0 , getObjValue());
makeBranch( e1 >= 1, getObjValue());

e1.end();

before this set up I also had it set up using range array which gave me a memory leakage at the IloRangeArray portion
IloRangeArray branches(getEnv());
branches.add(e1 == 0);
branches.add(e1 >= 1);

makeBranch( branches[0], getObjValue());
makeBranch( branches[1], getObjValue());
e1.end();

branches.end();
Thanks in advance for any help
Cam
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: EdKlotz

> clemsonmacrone wrote:
> Hi, I am having a memory leak at the makeBranch() lines of my branch callback routine. Is their something fundamentally wrong with my set up that causes this? note: F2Arcs is a vector of the variables I am branching on, and vars_ is a pointer to the variables in the model
>
>
> IloExpr e1(getEnv());
>
> for(int i =0; i < F2Arcs.size(); i++)
> {
> e1 += (*vars_)[F2Arcsi];
> }
>
> makeBranch( e1 == 0 , getObjValue());
> makeBranch( e1 >= 1, getObjValue());
>
> e1.end();

Keep in mind that the +=, == and >= operators here are actually functions
associated with these overloaded operator. I think you are OK regarding
the use of += in the above code; the results of that function call should
be updated into the existing IloExpr e1. But, I think your makebranch
calls create two additional IloRanges:

e1 == 0
e1 >= 1
>
>
>
> before this set up I also had it set up using range array which gave me a memory leakage at the IloRangeArray portion
>
>
> IloRangeArray branches(getEnv());
> branches.add(e1 == 0);
> branches.add(e1 >= 1);
>
> makeBranch( branches[0], getObjValue());
> makeBranch( branches[1], getObjValue());
> e1.end();
>
> branches.end();
>
>
> Thanks in advance for any help
> Cam

This second approach is on the right track. But, keep in mind that
branches is an IloRangeArray containing handles to the individual
IloRanges. So, while the above code frees the memory associated with
the array of handles, it doesn't free the memory associated with the
IloRanges to which the handles point to. With that in mind, try
preceding the call to branches.end() with

branches[0].end(); branches[1].end();

I hope that takes care of any leakage. If not, try systematically simplifying
your branch callback to help isolate the source of the leakage. For example,
try making e1 an expression consisting of just a single variable and comment
out the for loop.

For additional information regarding the creation and deletion of extractables,
look in the corresponding subsections of the Concepts section of the CPLEX C++
Reference Manual. Also, some general debugging tips that may help appear at
http://www-01.ibm.com/support/docview.wss?uid=swg21400038

Ed
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: clemsonmacrone

branches[1].end();
branches[0].end();
branches.end();

got rid of the leak
Thanks!
#CPLEXOptimizers
#DecisionOptimization