Hello,

I work on a particle system solver that uses conjugate gradient on large model cases. I wanted to try with Eigen's conjugate gradient to see if performances would be improved. The problem is that CG does not converge on a correct solution and always reaches the maximum number of iterations ( I tried with different limits ).

Is this a recurring problem ? I am probably doing something wrong... The Eigen version used is 3.3.2.

Here is how I proceed :

Code: Select all

SparseMatrix<double> A;
VectorXd b, x;

// A and b filled here

ConjugateGradient<SparseMatrix<double>, Lower|Upper> CGEigen;
CGEigen.setTolerance(tol); // tol = 1e-8
CGEigen.setMaxIterations(n); // I have tried multiple n values
CGEigen.compute(A);

x = CGEigen.solve(b);
Index iter = CGEigen.iterations();  // iter is always equal to n


Have a nice day!