[Coin-discuss] AIX Ampl->nl->Linux->cbc->sol->AIX Ampl not working

[Derek Jones]

Dear Andreas,

Thanks for the suggestion. I added that in and reran.

I got an ascii file, and, yes, a .sol file on AIX.

I copied the ASCII .nl file over to Linux, same problem.

:-(

I should note that the writesol.c funcs were writing in ASCII before. 
I'd already followed that through code tracing it.

But, yes, before ampl was indeed writing a binary .nl file.

However, doing the ASCII one didn't seem to make a difference sad to say.

Kind regards

Derek.

Andreas Waechter wrote:
> Hi,
> 
> One thing that could help is to generate the .nl file in ASCII format 
> and not as binary file (if you are not doing that already).  You do that 
> with
> 
> write gfilename;
> 
> to write the filename.nl file.
> 
> Maybe that will also instruct the solver executable to write the .sol 
> file as ASCII.
> 
> Andreas
> 
> On Fri, 22 Sep 2006, Kendall Bailey wrote:
> 
>> I have done almost exactly what you are describing.  I'm currently
>> using CPLEX as the solver on Linux ( AMD Opteron hardware ).  However,
>> I have done the same thing with an AMPL solver I wrote which utilizes
>> the Osi interfaces, and thus Cbc, Clp, Symphony, glpk, ... to do the
>> actual solve step.  I've had no endian troubles at all.  I'm using
>> AMPL on AIX just as you are.  Sorry I can't help with the Cbc program
>> itself.  I've planned to release the solver driver I wrote for some
>> time, but priorities changed.  I can send it to you if you'd like to
>> see it.
>>
>> Kendall
>>
>>
>> On 9/22/06, Derek Jones <scunacc at yahoo.com> wrote:
>>> Dear folks,
>>>
>>> What I'm trying to achieve may be a non-started because of endian /
>>> alignment issues, but I'm not sure.
>>>
>>> I want to do this:
>>>
>>> Have AMPL generate a .nl file, call a wrapper program instead of the
>>> solver, send the .nl file via the wrapper to Linux, where I will solve
>>> the problem using cbc file.nl -AMPL, then get the .sol file back on AIX.
>>>
>>> I have problems.
>>>
>>> Linux can't seem to write the .sol file properly. It *appears* to solve
>>> OK, but then bombs in the writesol.c file in getsufhead.
>>>
>>> I've traced thru, and there appears to be unreadable pointer memory
>>> inside the current "d" variable.
>>>
>>> Anyhow, has anyone got the ampl solver stuff to work OK on Linux? If so,
>>> what compile options did you use?
>>>
>>> Has anyone faced the above problem?
>>>
>>> Has anyone else :-) tried anything as crazy as the subject line at all ?
>>>
>>> Kind regards
>>>
>>> Derek Jones.
>>> _______________________________________________
>>> Coin-discuss mailing list
>>> Coin-discuss at list.coin-or.org
>>> http://list.coin-or.org/mailman/listinfo/coin-discuss
>>>
>> _______________________________________________
>> Coin-discuss mailing list
>> Coin-discuss at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/coin-discuss
>>
>