Hi Cherif, > > > The output you asked for is in the last mail. > > Thank you very much. > > Round2n works correctly, so most probably the error appears because the > current basis matrix becomes singular or ill-conditioned, and the > factorizer doesn't detect this. I cannot reproduce this failure because > failures of this kind are highly sensitive to round-off errors. I need a > time to prepare an appropriate test and hope you could help to run it. >
I could reproduce the failure. As I expected, the error appears in the sparse phase of basis factorization. Starting from some elimination step elements of the active submatrix grow catastrophically that eventually leads to nan in solution vector. Though the basis matrix seems to be close to singular, the factorizer doesn't detect this because all pivots have good magnitude. Thank you for your help, Andrew Makhorin _______________________________________________ Bug-glpk mailing list [email protected] https://lists.gnu.org/mailman/listinfo/bug-glpk
