[Ipopt] Slow execution

Andreas Waechter andreasw at watson.ibm.com
Mon Apr 13 12:17:35 EDT 2009

Hi Krish,

To follow up with Stefan's suggestion, if you use the option

mehrotra_algorithm yes

it chooses the adaptive barrier parameter option and a few other ones that 
might make things more efficient for LPs and convex QPs.

Also, in case you have relatively dense matrices, it might be worthwhile 
to try to use a good BLAS implementation (not netlib default code).



On Thu, 9 Apr 2009, Stefan Vigerske wrote:

> Hi,
>> I have a quadratic program (about 500 variables, 5 constraints, 500 lower and upper bounds).  I am using Mumps because of licensing issues.  I find that the program executes rather slowly ( takes about 2 to three times longer) than a commercial solver.  Are there any tweaks that I can do to improve performance?  At present I am using the default settings for the optimizer and have not chosen to override any of the settings in my own calling program.
> Setting mu_strategy to adaptive is sometimes helpful.
> Stefan
> -- 
> Stefan Vigerske
> Humboldt University Berlin, Numerical Mathematics
> http://www.math.hu-berlin.de/~stefan
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt

More information about the Ipopt mailing list