[SCIP] Inquiry Regarding SCIP Solver Performance with Parallel Option

[Stefan Vigerske]

Hi,

if the default settings are already "best" for the instance, then 
running additional solves with different settings in parallel 
(concurrently) won't improve performance but slow down also the first 
solve due to the machine being more busy.
Further, the concurrent mode is deterministic in default setting, which 
can mean that threads/solves have to wait for each other at 
synchronization points.

Stefan

On 02/12/2024 08:50, xu liwei wrote:
> Hi,
> I have been testing various MILP models with the SCIP solver and noticed that enabling the parallel option (using "parallel/maxnthreads" 4) results in slower performance on my 8-core CPU, which seems counterintuitive. In most cases, the parallel execution mode appears to decrease the computation speed rather than improve it.
> I am currently using the pre-compiled version of SCIP on a Windows platform and would appreciate any guidance or suggestions you might have regarding this issue. Could it be related to a misconfiguration, or are there other factors that might explain this behavior?
> Thank you very much for your time and assistance. I look forward to your insights.
> Best regards,
> Leslie
> 
> 
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip