[Coin-ipopt] using IPOPT to solve LPs

[Stefan Vigerske]

Hi,

I'm interesting in using IPOPT to solve (large) LPs.
Is there some non-default option-setting which might help to improve 
Ipopt's performance? Also experimentals options that I can play with are 
welcome :-).

On a first sight on the (long) list of Ipopt options, the following seem 
to be interesting:
dependency_detector
jac_c_constant
jac_d_constant
hessian_constant
mu_strategy
mehrotra_algorithm

I'm using Ipopt 3.3 with MA27 and MC19 via the GAMS interface.
(The GAMS interface tells Ipopt about variable/constraint-linearity via 
  TNLP::get_variables/constraints_linearity. Does it still benefit to 
set the jac_c/d_constant options?)

Best,
Stefan

-- 
Stefan Vigerske
Humboldt University Berlin, Numerical Mathematics
http://www.math.hu-berlin.de/~stefan