[Coin-lpsolver] quadratic objective

[Matthew Guthaus]

I know there's been messages on this before, but I'm confused what  
works, what is implemented but not tested, etc.

As far as I can tell, there are several methods to solve QP objective  
problems:

ONE WAY

solvectl.setSolveType(ClpSolve::useDual); // very fast, but  
getObjValue is always 0
-OR-
solvectl.setSolveType(ClpSolve::usePrimal); // slower, but  
getObjValue looks reasonable (different than CPLEX)
-OR-
solvectl.setSolveType(ClpSolve::useBarrier); // very slow
solvectl.setPresolveType(ClpSolve::presolveOn);
model.initialSolve(solvectl);

ANOTHER WAY
model.nonlinearSLP(100,1.0e-4); //


However, none of these methods get the same cost solution as CPLEX...  
what is the preferred method?

Thanks,

Matt