[Ipopt-tickets] [Ipopt] #242: High computation time in IPOPT

[Ipopt]

#242: High computation time in IPOPT
--------------------------+------------------------
Reporter:  rikrdos6       |      Owner:  ipopt-team
    Type:  clarification  |     Status:  new
Priority:  normal         |  Component:  Ipopt
 Version:  3.11           |   Severity:  normal
Keywords:                 |
--------------------------+------------------------
 Hello, I'm trying to solve a large sparse NLP problem with IPOPT via
 matlab, the problem has 70405 variables(It's a full discretization of a
 PDE system), when I solve the problem this is the output:

 {{{
 iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du
 alpha_pr  ls
    0 -9.0000000e+001 9.59e-012 2.33e-001   0.0 0.00e+000    -  0.00e+000
 0.00e+000   0
    1 -9.0009660e+001 1.50e-004 1.73e-002  -0.3 7.23e-002    -  9.94e-001
 1.00e+000f  1
    2 -9.0026898e+001 7.32e-005 3.57e-002  -6.0 1.67e-002   0.0 9.89e-001
 1.00e+000h  1
 Reallocating memory for MA57: lfact (182789633)
    3 -9.0078701e+001 6.58e-004 1.68e-002  -4.1 5.04e-002  -0.5 9.95e-001
 1.00e+000h  1
    4 -9.4061291e+001 1.57e+000 8.51e-001  -3.6 5.43e+002    -  2.80e-002 1
 .71e-002f  1
    5 -9.6071087e+001 2.90e+000 1.01e+000  -2.4 1.54e+002    -  1.00e+000 2
 .20e-002F  1
    6 -1.0613177e+002 4.31e+001 4.61e+000  -2.9 9.86e+001    -  6.06e-002 1
 .69e-001f  1
    7 -1.0553178e+002 5.36e+001 3.32e+002  -0.5 3.67e+001    -  1.10e-003 2
 .89e-001F  1
    8 -1.0580736e+002 5.32e+001 3.04e+002  -2.4 1.32e+002    -  1.77e-001 7
 .74e-003h  1
    9 -1.0678076e+002 5.09e+001 2.04e+002  -2.4 9.27e+001    -  1.46e-001 4
 .13e-002h  1
 }}}

 some time later...


 {{{
   35 -1.1419271e+002 1.14e+001 1.40e+002  -1.1 3.99e+001    -  1.91e-001 1
 .21e-001h  1
   36 -1.1445867e+002 5.28e+000 2.49e+002  -1.9 1.71e+001    -  1.00e+000 9
 .97e-001h  1
   37 -1.1429426e+002 1.41e-001 1.04e+001  -2.3 1.29e+000    -  1.00e+000 9
 .92e-001h  1
   38 -1.1434439e+002 2.17e-002 2.75e+000  -3.0 2.25e+000    -  1.00e+000 8
 .35e-001h  1
   39 -1.1435581e+002 7.07e-003 4.98e-001  -4.4 7.19e-001    -  9.99e-001 7
 .83e-001h  1
 iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du
 alpha_pr  ls
   40 -1.1435527e+002 2.00e-004 2.84e+000  -4.6 7.32e-003    -  1.00e+000 9
 .86e-001h  1
   41 -1.1435540e+002 1.60e-004 2.21e+001  -6.2 8.52e-003    -  1.00e+000
 1.00e+000h  1
   42 -1.1435537e+002 5.80e-006 9.62e-001  -7.4 1.18e-003    -  1.00e+000
 1.00e+000h  1
   43 -1.1435535e+002 1.26e-006 6.05e-001  -9.1 7.60e-005    -  1.00e+000
 1.00e+000h  1
   44 -1.1435536e+002 7.23e-009 1.43e-005 -10.6 2.08e-005    -  1.00e+000
 1.00e+000h  1

 Number of Iterations....: 44

                                    (scaled)                 (unscaled)
 Objective...............: -1.1435535863297133e+002
 -1.1435535863297133e+002
 Dual infeasibility......:  1.4319931187944945e-005
 5.8144724389685925e-006
 Constraint violation....:  7.2310643148189996e-009
 7.2310643148189996e-009
 Complementarity.........:  3.3012137777981496e-011
 3.3012137777981496e-011
 Overall NLP error.......:  1.4319931187944945e-005
 5.8144724389685925e-006


 Number of objective function evaluations             = 48
 Number of objective gradient evaluations             = 45
 Number of equality constraint evaluations            = 50
 Number of inequality constraint evaluations          = 50
 Number of equality constraint Jacobian evaluations   = 45
 Number of inequality constraint Jacobian evaluations = 45
 Number of Lagrangian Hessian evaluations             = 0
 Total CPU secs in IPOPT (w/o function evaluations)   =   2562.922
 Total CPU secs in NLP function evaluations           =     44.984
 }}}

 IPOPT is taking quite a while to solve the problem, and it is really
 important that the optimizer solves the problem in the least time
 possible, should I consider this normal? or do you have any suggestion of
 what could be wrong? I'm providing gradient and jacobian and using the
 quasi-newton approximation for the hessian.

--
Ticket URL: <https://projects.coin-or.org/Ipopt/ticket/242>
Ipopt <http://projects.coin-or.org/Ipopt>
Interior-point optimizer for nonlinear programs.