CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Preconditioned Gmres (

carlos April 8, 2003 06:29

Preconditioned Gmres
Hi all,

I have found a restarted gmres subroutine gmresr(j,m). It works well, but for my Navier-Stokes system [velocity + pressure] it is a little bit slow.

Can anyone answer me one of these two questions?

1- What the hell is the second parameter in the 'meaning' gmresr(j,m). I presume that the first is the krylov dimension (gmres will work with the "j" last vectors), but m???

2- If I wish to preconditionate the whole system, What preconditioner works? I have think in [Diagonal of velocity ] [ Lumped mass pressure] In fact, I have see that this preconditionner is utilised for the Evolutive problem, but for the stationnary problem is fine also??

Does anyone have the preconditioned subroutine for gmresr or boicgstab(4) (I mean, with the calling to the preconditioning in the routine)?

Thanks in advance,


kenn April 14, 2003 03:21

Re: Preconditioned Gmres
GMRES is a little slow but very stable.

BiCGSTAB is faster but less robust.

I am very interested in the preconditioned GMRES(l,m) you are talking about+

carlos April 14, 2003 05:24

Re: Preconditioned Gmres
I have NOT a preconditioned Gmres. That is precisely what I want. This is only the gmresr(l,m) in which it is necessary to introduce (=> modify) the preconditioning part.

All times are GMT -4. The time now is 18:39.