Home » Forums » METIS » Parallel Graph & Mesh Partitioning

Mc_SerialKWayAdaptRefine issue

I seem to be getting strange behavior from a call to IsHBalanceBetterTT in Mc_SerialKWayAdaptRefine. I believe the problem comes from the passing of the second argument. Inside IsHBalanceBetterTT, I see that pt1[1] is nan. The argument in IsHBalanceBetterTT looks suspect to me.

‹ Terminology and using weights Mc_SerialKWayAdaptRefine/IsHBalanceBetterTT ›
Submitted by jmousel on Tue, 2013-03-26 13:47
»

RE: I agree

That looks wrong to me as well.

Try the following:

IsHBalanceBetterTT(ncon, npwgts+mynbrs[k].pid*ncon, npwgts+to*ncon,nvwgt,ubvec);
^^^

and let me know if that works.

george

Submitted by karypis on Wed, 2013-03-27 16:44.
»