> El 13 jul 2022, a las 7:59, Runfeng Jin <[email protected]> escribió:
>
> Thank you, I will have a try to them.
>
> I use JD and GD because this matrix is about quantum chemical computing, and
> it has the property that diagonal elements dominate. This property seems to
> be suitable for JG and GD.
This is useful if you want to use a diagonal preconditioner (classical
Davidson), but generalized Davidson can take any preconditioner. In PETSc the
default preconditioner is not Jacobi but Block Jacobi+ILU.
>
> > JD and GD are best when you need to compute eigenvalues around a target
> About this, I want to compute only three smallest real eigenvalues, should
> this be the situation suitable for GD and JD？
No, try Krylov-Schur or LOBPCG.
>
> Runfeng
>
> Jose E. Roman <[email protected]> 于2022年7月13日周三 12:36写道：
> Does it work with -eps_gd_blocksize 1 ?
> Why do you want to use GD? JD and GD are best when you need to compute
> eigenvalues around a target. For your case, I would try with the default
> solver (Krylov-Schur) or with LOBPCG.
>
> Jose
>
>
> > El 13 jul 2022, a las 4:59, Runfeng Jin <[email protected]> escribió:
> >
> > Hi!
> > I am trying to find 3 eigenvalues of a matrix 81,160*81,160, and it get 0
> > eigenvalue after 162320 iterations. I use -eps_monitor and find error
> > estimation floats over 1e-6, I didn't set the tolerance so it should be the
> > default 1e-8. Actually it should achieve 1e-10.
> >
> > I have tried JD and GD, both of them can not converge. Does there any ways
> > to help it converge to the 1e-10?
> >
> > By the way, I use default Convergence criterion, the output of -log_view
> > and -eps_monitor are shown in attachment.
> > Thank you!
> >
> > Runfeng Jin
> > <eps_monitor-and-log_view.txt>
>
