Was wondering if any thought has been given to implementing a TFQMR solver with
CUSP.
Also, it would be useful if another Krylov solver criteria could be "minimum
iterations." In the research I've been working on (CFD) we have some
conditions (ie Steady/Laminar Flow) where the tolerance criteria is already met
for the current time step, but we still wish to apply the the solver, thereby
automatically providing us with higher accuracy.
Original issue reported on code.google.com by greg.j.v...@gmail.com on 24 Oct 2012 at 2:38
Original issue reported on code.google.com by
greg.j.v...@gmail.com
on 24 Oct 2012 at 2:38