[Ipopt] Running Time comparison of IPOPT Between MATLAB and C++

[Tony Kelman]

Robin,

This can vary substantially depending on the size and nonlinearity structure of your problem, and your Matlab coding methods. The most important information to look at is the timing results of “CPU secs in IPOPT” vs “CPU secs in NLP function evaluations.” The former depends on the number of iterations, algorithm options, and choice of linear system solver (Mumps vs MA57 etc) used by Ipopt, and you won’t be able to improve it much by coding your problem in C/C++. You can get more speedup in the NLP function evaluations, which is the time spent evaluating the Matlab objective, gradient, constraint, Jacobian and Hessian functions.

You can get a more detailed breakdown of Ipopt timing by setting the option print_timing_statistics to yes, and for the Matlab functions I highly recommend using the Matlab profiler to identify time-consuming parts of your (or COBRA’s) code.

-Tony
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20131206/321bb773/attachment.html>