[Couenne] Couenne install : problem with make thirdParty blas

Fri Oct 23 11:57:22 EDT 2009

Hello David,

> I updated my couenne from stable 0.2 and since then, I'm able to launch
> testcouenne.nl with no segmentation fault . I just have numerous messages
> such as :
>
> IpIpoptApplication.cpp" at line 725:
>  Exception message: status != TOO_FEW_DEGREES_OF_FREEDOM evaluated false:
> Too few degrees of freedom (rethrown)!

This message is printed by Ipopt, the NLP solver used by Couenne to  
find local optima of the continuous relaxation. The continuous NLP  
relaxation is causing some trouble, I guess the constraints make it  
difficult to find an initial solution. That should not be a big  
problem for Couenne, as feasible solutions are sought in many  
Branch-and-bound nodes and not only through Ipopt.

> An other message " You are using Ipopt by default with the MUMPS linear
> solver Other linear solvers might be more efficient (see Ipopt
> documentation)". I'll certainly have a look at the doc but would be Coin/Clp
> available to replace MUMPS, and if yes, would it be a benefit ?

I believe "linear solvers" here stands for "solvers of systems of  
linear equations" rather than Linear Programming solvers. An  
alternative to Mumps would be the HSL library, for example.

> For the blend.mod, I still have the segmentation fault, and nothing else
> even if I increase the verbose. However, I'm more interested in
> testCouenne.nl (my model) rather than blend.nl (AMPL NL sample file )

Given that I can't replicate that segfault, it would be useful if you  
send me the .mod file, the AMPL version you used, and the options you  
set, if any. Do you generate the .nl file on a Windows version of AMPL  
and then solve it using Couenne from Linux?

Cheers,
Pietro

PS: I sent a copy of this to the Couenne mailing list. In the future,  
I'd appreciate if you email all questions and comments to the list.

_________________________________________
Pietro Belotti, Lehigh University
Dept. of Industrial & Systems Engineering
200 W Packer Ave, Bethlehem PA 18015.
phone: 610-758-3865   fax: 610-758-4886
email: belotti at lehigh.edu
web:   http://www.lehigh.edu/~pib208

On 10/22/2009, David Gravot <dgravot at noos.fr> wrote:

> Hello Pietro,
>
> I updated my couenne from stable 0.2 and since then, I'm able to launch
> testcouenne.nl with no segmentation fault . I just have numerous messages
> such as :
>
> IpIpoptApplication.cpp" at line 725:
>  Exception message: status != TOO_FEW_DEGREES_OF_FREEDOM evaluated false:
> Too few degrees of freedom (rethrown)!
>
> An other message " You are using Ipopt by default with the MUMPS linear
> solver Other linear solvers might be more efficient (see Ipopt
> documentation)". I'll certainly have a look at the doc but would be Coin/Clp
> available to replace MUMPS, and if yes, would it be a benefit ?
>
> For the blend.mod, I still have the segmentation fault, and nothing else
> even if I increase the verbose. However, I'm more interested in
> testCouenne.nl (my model) rather than blend.nl (AMPL NL sample file )
>
> Many thanks for your help
>
> David
>
> -----Original Message-----
> From: couenne-bounces at list.coin-or.org
> [mailto:couenne-bounces at list.coin-or.org] On Behalf Of Pietro Belotti
> Sent: 22 October 2009 16:44
> To: couenne at list.coin-or.org
> Subject: Re: [Couenne] Couenne install : problem with make thirdParty blas
>
> Hello David,
>
> I could replicate the problem with testcouenne.nl, but not with
> blend.nl, which both the stable and trunk version seem to solve in a
> matter of seconds (optimum: 0.05172771358573224). The segfault with
> testcouenne.nl seems to be related to four redundant constraints in
> the model that look as follows:
>
> -x26+x25=0
> +x26-x25=0
> -x26+x25=0
> +x26-x25=0
>
> I have just updated both the stable and the trunk version, and the
> problem seems to go away (although testcouenne.nl seems quite
> difficult). As for blend.nl, can you tell when the segfault occurs, or
> can you send the output you get? It would be especially useful if you
> set the problem_print_level to 7 in the couenne.opt option file.
>
> Cheers,
> Pietro
>
> _________________________________________
> Pietro Belotti, Lehigh University
> Dept. of Industrial & Systems Engineering
> 200 W Packer Ave, Bethlehem PA 18015.
> phone: 610-758-3865   fax: 610-758-4886
> email: belotti at lehigh.edu
> web:   http://www.lehigh.edu/~pib208
>
>
> On 10/21/2009, David Gravot <dgravot at noos.fr> wrote:
>
>> Hello
>>
>> Indeed, I'm now able to launch the make unitTest on the new ubuntu
> release.
>> However, I have generated a testcouenne.nl file using AMPL that seems to
>> generate a segmentation fault once I launch couenne (although couenne
>> correctly displays infos on the model). Same segmentation fault if I
>> generate blend.nl file from my ampl Non-linear models directory.
>>
>> I must say that I'm generating this file through AMPL on a windows
>> environment. Could it create a problem ?
>>
>> Regards
>>
>> David
>>
>> -----Original Message-----
>> From: couenne-bounces at list.coin-or.org
>> [mailto:couenne-bounces at list.coin-or.org] On Behalf Of Pietro Belotti
>> Sent: 07 October 2009 05:07
>> To: couenne at list.coin-or.org
>> Subject: Re: [Couenne] Couenne install : problem with make thirdParty blas
>>
>> Hello David,
>>
>> it seems related to the use of fort77 as Fortran compiler. If you have
>> other compilers, such as gfortran, can you re-issue the configure
>> command with option "F77=gfortran" and see if the problem persists? If
>> this fails, please send what type of machine you are using and what
>> version of Couenne you have downloaded (stable? trunk?).
>>
>> Best regards,
>> Pietro
>>
>>
>> _________________________________________
>> Pietro Belotti, Lehigh University
>> Dept. of Industrial & Systems Engineering
>> 200 W Packer Ave, Bethlehem PA 18015.
>> phone: 610-758-3865   fax: 610-758-4886
>> email: belotti at lehigh.edu
>> web:   http://www.lehigh.edu/~pib208
>>
>>
>> On 10/06/2009, dgravot at noos.fr wrote:
>>
>>      Hello
>>
>> I'm facing a problem during make invoking third party BLAS.
>> The configure phase however ran correctly (configure: Configuration
>> of Couenne successful
>> configure: Main configuration of Couenne successful). The install for
>> BLAS ran also fine. Looking at the trace, it seems that dasum.lo has
>> been generated by fort77 instead of dasum.o ?
>>
>> Making all in ThirdParty/Blas
>> make[1]: entrant dans le répertoire «
>> /home/dgravot/Couenne/build/ThirdParty/Blas »
>> /bin/bash ./../../libtool --tag=F77 --mode=compile fort77  -O3
>> -fomit-frame-pointer -pipe   -c -o dasum.lo
>> ../../../ThirdParty/Blas/dasum.f
>>   fort77 -O3 -fomit-frame-pointer -pipe -c
>> ../../../ThirdParty/Blas/dasum.f  -fPIC
>>     dasum:
>> mv -f dasum.o .libs/dasum.o
>> mv: cannot stat `dasum.o': No such file or directory
>> make[1]: *** [dasum.lo] Erreur 1
>> make[1]: quittant le répertoire «
>> /home/dgravot/Couenne/build/ThirdParty/Blas »
>> make: *** [all-recursive] Erreur 1
>>
>> Thanks for your help
>>
>> David
>>
>>
>>
>>
>> ----------------------------------------------------------------
>> This message was sent using IMP, the Internet Messaging Program.
>>
>>
>> _______________________________________________
>> Couenne mailing list
>> Couenne at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/couenne
>>
>> __________ NOD32 4485 (20091006) Information __________
>>
>> This message was checked by NOD32 antivirus system.
>> http://www.eset.com
>>
>>
>
>
>
> ----------------------------------------------------------------
> This message was sent using IMP, the Internet Messaging Program.
>
>
> _______________________________________________
> Couenne mailing list
> Couenne at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/couenne
>
> __________ NOD32 4533 (20091022) Information __________
>
> This message was checked by NOD32 antivirus system.
> http://www.eset.com
>
>

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.