Newsgroups: comp.parallel.mpi
From: Guillaume Dumont <Guillaume.Dumont@inria.fr>
Subject: PETSC : Questions about iterations
Organization: INRIA Rocquencourt
Date: Mon, 17 Aug 1998 18:49:02 +0200
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <35D85EFE.167E@inria.fr>

I am using PETSC with SLES to solve many complex non-hermitian systems.
When I use KSP_TYPE=GMRES(30) and PC_TYPE=BJACOBI I obtain really
different number of iterations between odd and even number of procs.

NPROC  TIME(s) ITERATIONS 
1	241	554
2	185	1085
3	93	579
4	85	1089
5	57	592
6	59	1093
7	45	610
8	51	1095

Could anyone help me please. Thanks.

Guillaume DUMONT