Hi Francois, > I am trying to benchmark ViennaCL against the MKL for GMRES > preconditioned by ILU0 factorization. It turns out that I get different > number of iterations for both solvers. Where can I get more information > about the convergence criteria? > > More precisely, if we are trying to solve A.x = y, preconditioned by B > (so we try to solve B.A.x = B.y If I understand correctly), with a > relative error of epsilon, what is the convergence criteria ? > - Is it Norm(A.x_n - y) <= epsilon Norm(y) ? > - Which norm is used ? > - Can we customize the norm used?
The convergence criterion is monitoring the relative preconditioned 2-norm, i.e. ||BAx - By|| // ||By|| You can customize the termination criterion by passing a suitable monitor routine, but you cannot modify the norm used for monitoring convergence. Best regards, Karli > > Best regards, > > François Fayard > Founder & Consultant - Inside Loop > Tel: +33 (0)6 01 44 06 93 > Web: www.insideloop.io <http://www.insideloop.io> > > > > ------------------------------------------------------------------------------ > Developer Access Program for Intel Xeon Phi Processors > Access to Intel Xeon Phi processor-based developer platforms. > With one year of Intel Parallel Studio XE. > Training and support from Colfax. > Order your platform today.http://sdm.link/xeonphi > > > > _______________________________________________ > ViennaCL-support mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/viennacl-support > ------------------------------------------------------------------------------ Developer Access Program for Intel Xeon Phi Processors Access to Intel Xeon Phi processor-based developer platforms. With one year of Intel Parallel Studio XE. Training and support from Colfax. Order your platform today.http://sdm.link/xeonphi _______________________________________________ ViennaCL-support mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/viennacl-support
