From sistek at math.cas.cz  Mon Oct  9 22:41:21 2017
From: sistek at math.cas.cz (Jakub Sistek)
Date: Mon, 9 Oct 2017 21:41:21 +0100
Subject: [bddcml-users] Question about Multilevel BDDC solver library
In-Reply-To: <CANZZEumO=JSFP-+5yogX1s+zopykTCo5qp4VO9r+4Er-SBDykA@mail.gmail.com>
References: <CANZZEumO=JSFP-+5yogX1s+zopykTCo5qp4VO9r+4Er-SBDykA@mail.gmail.com>
Message-ID: <67dc82c5-a9c4-4e13-fa80-2fed06400007@math.cas.cz>

Dear Jianrong,

thank you for you interest in BDDCML.

I would like to encourage you to start playing with and looking into the 
poisson_on_cube example, which does not require any input files. The 
other examples are somewhat obsolete. This example solves a Poisson 
problem on a unit cube. It has three arguments - number of elements per 
an edge of a cubic subdomain, number of such subdomains along an edge of 
the unit cube, and the number of levels in BDDC.
For example,

mpirun -np 2 ./poisson_on_cube 4 2 2

runs the example using 2 MPI processes, with subdomains of 4x4x4 
trilinear elements, using 2x2x2=8 subdomains, and 2 levels in BDDC (the 
standard BDDC method with a coarse problem). The solution can be 
examined from the created pvd file.

The example is newer and more self-contained than bddcml_{local, global}.

I am very interested in your application, so please, do not hesitate to 
ask if needed.

Best wishes,

Jakub



On 9.10.2017 17:48, Jianrong Lu wrote:
> Hi Dr. Sistek
>                I am Jianrong. I am  working on high performance 
> computing. I try to solve fluid structure interaction problem. I try 
> to test bddcml before I use bddcml to solve my question.
>               Would you mind to give me a example how to describe 
> 'PROBLEMNAME' ? when I running
>
>               mpirun -np X ./bddcml_local PROBLEMNAME M
>
> Thank you.
>
> Jianrong