[SCIP] Multiple threads with SCIP and Google OR Tools

[Ines Spenke]

Hi Marc,

Thank you for answering and nice to hear from you, too!

We use SCIP via Google OR Tools. So, I cannot try SCIP 10 that easily before Google OR Tools is updated, too. That’s a pity.

Although there seems to be a bug, we see a much better performance when running the model on several threads. So, I am glad about this concurrent mode.

If you could profit from our experience with this bug situation or if you need someone to try out something, just don’t hesitate to contact. 

Best,
Ines

-----Ursprüngliche Nachricht-----
Von: Scip <scip-bounces at zib.de> Im Auftrag von Marc Pfetsch
Gesendet: Donnerstag, 11. Dezember 2025 19:22
An: scip at zib.de
Betreff: Re: [SCIP] Multiple threads with SCIP and Google OR Tools

[EXTERN!]

Hi Ines,

nice to hear from you!

Have you used SCIP 10?

There was a bug in SCIP 9 (and older) in the concurrent solve, which we
(partly) solved. It is on my (our) todo list to look into this in more detail.

Thus, you can try SCIP 10 and see whether the problems go away (if not done already). Or you do not use concurrent mode. I think that currently you will not loose too much performance when doing this.

Best

Marc


On 11/12/25 17:56, Ines Spenke wrote:
>
> Hi SCIP Team,
>
> I try to solve a MIP with SCIP within Google OR-Tools. It is set up 
> with 8 threads and a time limit of 90min. I do not know if the 
> following problem is due to Google OR or SCIP.
>
> In the console output, I see a progress table with dualbounds and 
> primalbounds as below.
>
> The final primal bound reported at termination (4.03e+04) is much 
> worse than the last values shown in the table (e.g. 2.50e+04). The 
> same for dual.
>
> This looks like SCIP found better solutions during the run, but they 
> are not kept until the end.
>
> Could you please help me to understand what exactly is shown in the 
> progress table and if there is a good reason to see better values 
> there than in the final solution?
>
> Thank you very much!
>
> Best regards,
>
> Ines
>
> time | mem | dualbound | primalbound | gap
>
> $ 506s|1407M| 6.131990e+03 | 3.899735e+09 | Large
>
> …
>
> $3186s|2181M| 6.582145e+03 | 4.037315e+04 | 513.37%
>
> …
>
> $5165s|3298M| 6.956456e+03 | 2.509565e+04 | 260.75%
>
> concurrent solver 'scip-2' stopped with status time limit reached
>
> concurrent solver 'scip-5' stopped with status time limit reached
>
> concurrent solver 'scip-3' stopped with status time limit reached
>
> concurrent solver 'scip-8' stopped with status time limit reached
>
> concurrent solver 'scip-6' stopped with status time limit reached
>
> concurrent solver 'scip-7' stopped with status time limit reached
>
> 5402s|3106M| 6.967374e+03 | 2.509565e+04 | 260.19%
>
> concurrent solver 'scip-1' stopped with status time limit reached
>
> concurrent solver 'scip-4' stopped with status time limit reached
>
> *5402s|3106M| 6.967374e+03 | 2.509565e+04 | 260.19% *
>
> SCIP Status : solving was interrupted [time limit reached]
>
> Solving Time (sec) : 5402.00
>
> Solving Nodes : 3941
>
> *Primal Bound : +4.03731513404846e+04 (44 solutions) *
>
> *Dual Bound : +6.24112774276733e+03 *
>
> Gap : 546.89 %
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
_______________________________________________
Scip mailing list
Scip at zib.de
https://listserv.zib.de/mailman/listinfo/scip