[Ipopt-tickets] [Ipopt] #114: Assertion failure: HaveAffineDeltas false in ComputeSearchDirection
Ipopt
coin-trac at coin-or.org
Wed Apr 28 11:51:13 EDT 2010
#114: Assertion failure: HaveAffineDeltas false in ComputeSearchDirection
------------------------------+---------------------------------------------
Reporter: guest | Owner: andreasw
Type: defect | Status: new
Priority: normal | Component: Ipopt
Version: 3.8 (C++ Version) | Severity: normal
Keywords: |
------------------------------+---------------------------------------------
Comment(by andreasw):
(this was sent to Ipopt-tickets at list.coin-or.org)
{{{
Hi,No problem about taking a while to reply - I guess you have many other
things to do.
I'm afraid I'm restricted in what I can say due to non-disclosure
restrictions, but I'll tell
you what I can, and what I can recall - we've been using a work-around for
some time, changing
line 82 of IpPDSearchDirCalc.cpp to check �HaveAffineDeltas(); my
understanding of what was in
that branch of the if statement was that it was an optimisation, so could
be excluded.
We have written a program which uses Ipopt from C++, so I can't give you
an AMPL example. The
problem in question is a quadratic problem with linear constraints, and we
use the mumps
solver. I believe it occurred when the problem happened to only have a
single feasible solution
- all the variables are constrained to zero.
I've set the following options:
jac_c_constant�yes
jac_d_constant�yes
hessian_constant�yes
mehrotra_algorithm�yes
and the following ones which I assume will make little difference:
print_level to 12
check_derivatives_for_naninf
derivative_test to second_order
print_timing_statistics yes
I hope this helps, if I can do anything else just drop me a mail - I can
reproduce it here,
Dave.
PS Are you aware that the "function evaluation" timing statistics appear
not to reset between
runs, whereas the others do, leading to some slightly confusing output.
}}}
--
Ticket URL: <https://projects.coin-or.org/Ipopt/ticket/114#comment:2>
Ipopt <http://projects.coin-or.org/Ipopt>
Interior-point optimizer for nonlinear programs.
More information about the Ipopt-tickets
mailing list