parallel ignores number of threads

[mhechthz]

With pyScipOpt 5.5.0 for a MILP problem I did set parallel parameters (I have 12 cores with 24 threads available):

        model.setParam("lp/threads", 8)  
        model.setParam("parallel/mode", 0) 
        model.setParam("parallel/minnthreads", 8) 
        model.setParam("parallel/maxnthreads", 8) 

But on running I only get 4 cores active working. Why? And why these busy not busy cycles? CPLEX uses all (!) threads.

[Joao-Dionisio]

Dear @mhechthz , SCIP is a sequential solver. Extensions for parallelizing it exist, like FiberSCIP. In the vanilla solver, the best bet is to compile it with the option -DTPI=tny, to enable solving the same problem in parallel with different settings and some communication between them. The method to optimize the problem is solveConcurrent().

I'm on mobile, apologies if there is poor formatting.