[Cbc] CBC significant different running times with the same instance and executable

[Ted Ralphs]

It can't find the solution because you imposed a time limit? Or there is an
error? In general, if the parallel version can find a solution, the
sequential version should also be able to find one, just in a (possibly
much) longer time frame.

Variability in running time is expected in parallel mode, since the
execution is non-deterministic by default. Cbc does have a deterministic
parallel mode, but imposing determinism has a cost. On average, the
non-deterministic version will be faster.

Ted

On Thu, Jan 30, 2020 at 11:52 AM Rita Portugal <rita at siscog.pt> wrote:

> Thank you for the quick answer.
>
> Unfortunally when turnning off parallel runnings CBC can't find the
> integer solution anymore.
>
>
> On 27/01/2020 22:17, Ted Ralphs wrote:
>
> Are you running in parallel or sequential mode? Parallel execution can be
> non-deterministic and may exhibit the variance you are seeing, depending on
> the parameters.
>
> Ted
>
> On Mon, Jan 27, 2020 at 12:10 PM Rita Portugal <rita at siscog.pt> wrote:
>
>> Hi,
>>
>> My name is Rita Portugal
>>
>> I've been working with CBC "inside" C++ language in Visual Studio
>> framework.
>>
>> I'm having some strange behavior that I'm hopping you could help me on
>> understanding it.
>>
>> I read the input data, create a linear integer model than call
>> branchAndBound.
>>
>> I started by testing a very small example to understand running time
>> behavior but, while running the same instance and the same solver
>> configuration several different times (in the same machine) I'm keep having
>> different time results for getting the same optimal solution (between 5
>> minutes to 25).
>>
>> When reading the user guide I understand that CBC has random components
>> during is running process that can be desactivate by the
>> setRandomSeed(7777) comand. But even doing that the behavior is still the
>> same.
>>
>> Can you please advise me about any other possible responsable components
>> for having such a different behavior between runnings?
>>
>> Thank you in advance
>> --
>>
>> Rita Portugal
>> Software Engineer
>>
>> *SISCOG - Sistemas Cognitivos, SA*
>> Campo Grande, 378 - 3º, 1700-097 Lisboa • Portugal
>> Tel: +351 217 529 100 • Fax: +351 217 529 101
>>
>>
>> *www.siscog.pt* <http://www.siscog.pt>
>>
>> *"Optimising the resources of the world"*
>>
>>
>>
>> DISCLAIMER
>> This message may contain confidential information. You should not copy or
>> address this message to third parties.
>> If you are not the appropriate recipient we kindly ask you to delete the
>> message and notify the sender.
>> The contents of this message and its attachments are the sole
>> responsibility of the sender and under no circumstances can SISCOG -
>> Sistemas Cognitivos, SA be liable for any resulting consequences.
>>
>> AVISO
>> Esta mensagem pode conter informação considerada confidencial, não
>> devendo ser copiada ou endereçada a terceiros.
>> Se o receptor não for o destinatário apropriado, agradecemos que destrua
>> a mensagem e informe o emissor do sucedido.
>> O conteúdo desta mensagem bem como dos respectivos anexos é da
>> responsabilidade exclusiva do emissor, não podendo a SISCOG - Sistemas
>> Cognitivos, SA ser responsabilizada por eventuais consequências.
>> _______________________________________________
>> Cbc mailing list
>> Cbc at list.coin-or.org
>> https://list.coin-or.org/mailman/listinfo/cbc
>>
>
>
> --
> Dr. Ted Ralphs
> Professor, Industrial and Systems Engineering
> Lehigh University
> (610) 628-1280
> ted 'at' lehigh 'dot' edu
> coral.ie.lehigh.edu/~ted
>
>
> --
>
> Rita Portugal
> Software Engineer
>
> *SISCOG - Sistemas Cognitivos, SA*
> Campo Grande, 378 - 3º, 1700-097 Lisboa • Portugal
> Tel: +351 217 529 100 • Fax: +351 217 529 101
>
>
> *www.siscog.pt* <http://www.siscog.pt>
>
> *"Optimising the resources of the world"*
>
>
>
> DISCLAIMER
> This message may contain confidential information. You should not copy or
> address this message to third parties.
> If you are not the appropriate recipient we kindly ask you to delete the
> message and notify the sender.
> The contents of this message and its attachments are the sole
> responsibility of the sender and under no circumstances can SISCOG -
> Sistemas Cognitivos, SA be liable for any resulting consequences.
>
> AVISO
> Esta mensagem pode conter informação considerada confidencial, não devendo
> ser copiada ou endereçada a terceiros.
> Se o receptor não for o destinatário apropriado, agradecemos que destrua a
> mensagem e informe o emissor do sucedido.
> O conteúdo desta mensagem bem como dos respectivos anexos é da
> responsabilidade exclusiva do emissor, não podendo a SISCOG - Sistemas
> Cognitivos, SA ser responsabilizada por eventuais consequências.
>


-- 
Dr. Ted Ralphs
Professor, Industrial and Systems Engineering
Lehigh University
(610) 628-1280
ted 'at' lehigh 'dot' edu
coral.ie.lehigh.edu/~ted
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20200130/7179aa0d/attachment.html>