[Coin-ipopt] CUTEr quesiton - C++ version

[Oliveira, Ivan B]

Hi Guys,

 

We're using the excellent C++ version of ipopt (3.0.0) which requires us
to now code our own quasi-Newton scheme.  We're noticing different
behavior for various combinations of BFGS and SR1 for our particular
small problem set so we wanted a more objective analysis.  We decided to
test our implementation on the CUTEr set, since this interface seems to
be supported with the ipopt distribution.  But we're having some
problems with ipopt-3.0.0/Apps/CUTErInterface.  First, we installed SIF
and CUTEr and tested it with 'sdgen KISSING2', which works as expected.
Then we installed ipopt with a pretty standard configuration, and tested
the library libipopt.a on several example problems.  Then, in
ipopt-3.0.0/Apps/CUTErInterface (we've already built ipopt), we:

 

> make (seems to work)

 

if test -z "$MYCUTER" -o -z "$CUTER"; then \

  touch ipoma.f; \

else \

  sed "/#{sizing}/r $MYCUTER/double/config/tools.siz" CUTErInterface.f |
sed -f $MYCUTER/double/config/cast.sed > ipoma.f; \fi

.../gcc/3.1/bin/g++   -c -o ipoma.o ipoma.f

rm -f libdummy.a

ar crv libdummy.a ipoma.o

a - ipoma.o

ranlib libdummy.a

 

> make cuter (seems to work)

 

sed -f $MYCUTER/double/config/script.sed ipo.pro > ipo

chmod a+x ipo

sed -f $MYCUTER/double/config/script.sed sdipo.pro > sdipo

chmod a+x sdipo

if test -z "$MYCUTER" -o -z "$CUTER"; then \

  echo CUTEr environment variables not set - cannot install ipoma.f; \

else \

  echo "-L.../ipopt-3.0.0/install/lib -lipopt  -lg2c"
>$MYCUTER/double/lib/ipo.liblist ;\

  cp ipoma.o .../CUTEr.large.pc.lnx.g77/double/bin/ipoma.o; \

  cp ipo .../CUTEr.large.pc.lnx.g77/bin/ipo ;\

  cp sdipo .../CUTEr.large.pc.lnx.g77/bin/sdipo ;\

fi

 

> ./sdipo KISSING2 

 

 Problem name: KISSING2

 

 Double precision version will be formed.

 

 The objective function uses        1 nonlinear group

 

 There are      625 nonlinear inequality constraints

 

 There are       94 free variables

 There are        6 fixed variables

 

 

.../CUTEr.large.pc.lnx.g77/double/bin/linpac.o: In function
`dcopy_':.../CUTEr.large.pc.lnx.g77/double/bin/linpac.o(.text+0x0):
multiple definition of `dcopy_'

.../ipopt-3.0.0/install/lib/libipopt.a(dcopy.o)(.text+0x0): first
defined here

/usr/bin/ld: Warning: size of symbol `dcopy_' changed from 618 to 373 in
.../CUTEr.large.pc.lnx.g77/double/bin/linpac.o

.../CUTEr.large.pc.lnx.g77/double/bin/linpac.o: In function `ddot_':

.../CUTEr.large.pc.lnx.g77/double/bin/linpac.o(.text+0x178): multiple
definition of `ddot_'

.../ipopt-3.0.0/install/lib/libipopt.a(ddot.o)(.text+0x0): first defined
here

/usr/bin/ld: Warning: size of symbol `ddot_' changed from 605 to 312 in
.../CUTEr.large.pc.lnx.g77/double/bin/linpac.o

 

And so on.  Seems there might be some doubly defined BLAS symbols in
libpac.o and libipopt.a?  Have you guys run into this problem before, or
are we doing something wrong?

 

--Ivan.

 

----------------------------

Ivan B. Oliveira

SC12-205

(408)765-0584

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/ipopt/attachments/20050908/251c00a7/attachment.html