[Ipopt-tickets] [Ipopt] #248: Solvers

[Ipopt]

#248: Solvers
----------------------------+------------------------
  Reporter:  mathieuDS      |      Owner:  ipopt-team
      Type:  clarification  |     Status:  closed
  Priority:  normal         |  Component:  Ipopt
   Version:  3.11           |   Severity:  normal
Resolution:  fixed          |   Keywords:
----------------------------+------------------------
Changes (by stefan):

 * status:  new => closed
 * resolution:   => fixed


Comment:

 Hi,

 Replying to [ticket:248 mathieuDS]:
 > I am trying to compile IPOPT with the solvers. There are several
 problems:
 > 1) The ./configure --help gives possible options:
 > {{{
 >   --with-ipopt-lib linker flags for using project Ipopt
 >   --with-ipopt-incdir directory with header files for using project
 Ipopt
 >   --with-ipopt-datadir directory with data files for using project Ipopt
 > }}}
 > which in my opinion does not make any sense since IPOPT should not
 depend on the location of IPOPT or any previous software.

 That comes with the build system and its distinction into base and project
 directory. I don't think it's worth to spend time on hiding these.

 > 2) It is possible to compile IPOPT without any solver, which means that
 later on the "make test" fails because M!27 is missing. The compilation
 should be stopped right when it is detected that some software are
 missing. Clear possible directions should be given in the "configure" so
 that we know what needs to be done.

 Ipopt can load some of the linear system dynamically at runtime, so
 building Ipopt without any linear solver is not an error per se.

 > 3) The information about the solver is dispersed in many different
 files. For example IpAlgBuilder.cpp lists the following: ma27, ma57, ma77,
 ma86, pardiso, ma97, wsmp, numps. But the "ThirdParty" directory lists:
 ASL, Blas, HSL, HSLold, Lapack, Metis, Mumps
 > It would be nice to know which correspond to which, relative
 performance, license, all in one single information file.

 The documentation gives some [http://www.coin-
 or.org/Ipopt/documentation/node13.html hints]. We surely don't want to do
 statements on performance of third-party codes.

--
Ticket URL: <https://projects.coin-or.org/Ipopt/ticket/248#comment:1>
Ipopt <http://projects.coin-or.org/Ipopt>
Interior-point optimizer for nonlinear programs.