Florian Pommerening

Results 45 comments of Florian Pommerening

High runtime of `_rationalLUSolver.clear()` in `SoPlexBase<R>::_changeLhsReal()`

The way I understood things, SoPlex doesn't directly implement primal/dual simplex but the the algorithm depends on whether the representation is in column or row form (https://soplex.zib.de/doc/html/FAQ.php#primaldualsimplex). Could this have...

High runtime of `_rationalLUSolver.clear()` in `SoPlexBase<R>::_changeLhsReal()`

I tried running the code with the 4 combinations of algorithm and representation. While the logs are different, all combinations had a larger number of iterations compared to the 3.1.1...

High runtime of `_rationalLUSolver.clear()` in `SoPlexBase<R>::_changeLhsReal()`

I did find a call to it on the second solve (`SPxSolverBase::solve` -> `SPxSolverBase::init` -> `SPxBasisBase::load` -> `SPxBasisBase::load` -> `SPxBasisBase::restoreInitialBasis`) and traced why the model was uninitialized. It turns out...

High runtime of `_rationalLUSolver.clear()` in `SoPlexBase<R>::_changeLhsReal()`

I didn't actively get or set the basis in the version using OSI but maybe this is something that OSI does internally? The profile I posted earlier shows some time...

High runtime of `_rationalLUSolver.clear()` in `SoPlexBase<R>::_changeLhsReal()`

Great to hear about the release. The experiments I ran unfortunately showed very mixed results. Sometimes the time for storing the basis just does not pay off as it did...