ortools linear_solver scip speed up - multithreading?

Question

I have the following code:

from ortools.linear_solver import pywraplp
SOLVER = pywraplp.Solver.CreateSolver('SCIP')
...
SOLVER.solve()

It executes for really long time - so to optimize it I attempted: SOLVER.SetNumThreads(8) prior to line with .solve() - but it didn't work at all then.

So I have the following questions:

I assumed multithreading to be a low-hanging fruit here, so how can I make it work - we're talking about regular python shell, in particular on jupyter notebook (but we can also just run this as python script.py - I just can't see reason why it wouldn't run on jupyter, but run as script).
Is there maybe any other way to speed it up? Maybe something less obvious?

Caveat I'm data engineer, and I assume whatever was done in there by data scientists is perfect - so I'm looking for recommendation how to work faster with ortools and not how to replace them with faster tool ;)

Laurent Perron · Accepted Answer · 2021-03-23 13:41:32Z

3

So,

multithread is not a low hanging fruit for SCIP. The best you can achieve is 2x.
It should be compiled in, except for linux, where for some reason, the cmake command to build it does not seem to work.
If your problem is purely integer, you can try the 'sat' solver. It benefits from multi-threading (8 workers or more).

answered Mar 23, 2021 at 13:41

Laurent Perron

11.2k1 gold badge10 silver badges27 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

ortools linear_solver scip speed up - multithreading?

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related