[Symphony] Symphony in a shared-memory architecture

[Ted Ralphs]

I have fixed a minor bug related to the Open MP version into both
trunk and stable version 5.3 that seems to fix the segmentation fault.
Can you check it and see if you still observe the same behavior? At
the moment, trunk seems to be broken for other reasons, so please
check out stable version 5.3 if you can.

Cheers,

Ted

On Wed, Feb 9, 2011 at 3:09 PM, Aiko Vogelsang <aikovogelsang at gmx.de> wrote:
> Hello everybody,
>
> at first thanks for the answer! The selection of the max. active nodes works
> fine.
>
> But with the parallelization another problem occurred. If I use Symphony in
> parallel I get a "Segmentation fault" after some calculations. If I use a
> sequential version of Symphony for exactly the same problem (the only
> difference in this case is "sym_set_int_param(env,  "max_active_nodes",
> 1);") everything looks like expected before (no failure).
>
> Secondly, if I get no segmentation fault (happens sometimes), the results of
> the calculation in parallel differ from the sequential case.
>
> Is it possible that there's a still a race condition between the threads? If
> so, I wouldn't understand it because I configured the solver for a parallel
> environment.
>
> Does anybody have an idea what can be wrong here or where I could have a
> look at?
>
> Again many thanks in advance for your support (I'm sorry to keep bothering
> you with my problem)!
>
> Best regards
> Aiko
>
>
> -----Ursprüngliche Nachricht-----
> Von: Ted Ralphs [mailto:ted at Lehigh.EDU]
> Gesendet: Montag, 7. Februar 2011 17:30
> An: Aiko Vogelsang
> Cc: symphony at list.coin-or.org
> Betreff: Re: [Symphony] Symphony in a shared-memory architecture
>
> The parameter "max_active_nodes" controls how many threads are used
> with OpenMP. To set that parameter through the API, do, e.g.,
>
> sym_set_int_param(env,  "max_active_nodes", 8);
>
> Let me know if that fixes your problem.
>
> Cheers,
>
> Ted
>
> On Thu, Feb 3, 2011 at 5:12 AM, Aiko Vogelsang <aikovogelsang at gmx.de> wrote:
>> Hello everybody,
>>
>> I've a question on how to use Symphony in a shared-memory architecture
> again.
>>
>> Because of the hints I got last month I compiled Symphony with the option
> "--enable-openmp". Everything works fine during compiling and installing.
>> If I invoke the solver from the command line with the option "-p 4"
> Symphony works in parallel (as I read in the manual).
>>
>> But if I include the header files and the library files of Symphony and
> the related projects into my C++ project the solver still works in
> sequential. I tried to insert some compiler (which is GCC 4.5) options but
> nothing worked.
>> According to my understanding I need an additional option compareable to
> "-p 4" that would be inserted in the command line.
>>
>> If I could get a hint for this problem it would be very helpful! Many
> thanks in advance!
>>
>> Best regards
>> Aiko
>>
>> _______________________________________________
>> Symphony mailing list
>> Symphony at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/symphony
>>
>
>
>
> --
> Dr. Ted Ralphs
> Associate Professor, Lehigh University
> (610) 628-1280
> ted 'at' lehigh 'dot' edu
> coral.ie.lehigh.edu/~ted
>
>
> _______________________________________________
> Symphony mailing list
> Symphony at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/symphony
>



-- 
Dr. Ted Ralphs
Associate Professor, Lehigh University
(610) 628-1280
ted 'at' lehigh 'dot' edu
coral.ie.lehigh.edu/~ted