[petsc-users] GAMG and linear elasticity

Tabrez Ali stali at geology.wisc.edu
Sun Sep 1 11:01:55 CDT 2013


Here's the code http://defmod.googlecode.com/

If you have PETSc installed with METIS then you can build the executable 
simply by running "make all PETSC_DIR=/home/user/petsc"

The three meshes that I am testing GAMG on are available at 

Here's how I run it:

$ for i in small.inp medium.inp large.inp ; do mpirun -np 4 ./defmod -f 
$i -pc_type gamg -ksp_monitor; done

and I get 16,28 & 52 iterations with ASM and 22,22 & 31 iterations with 
GAMG. A gif file showing the viz is attached. The problem simply 
involves application of a point force on a unit cube (nodes on the 
bottom and four side boundaries are fixed).

I have little idea about AMG and am experimenting/investigating if I can 
easily use it for some of my problems.


On 09/01/2013 10:44 AM, Mark F. Adams wrote:
> On Sep 1, 2013, at 8:37 AM, Tabrez Ali<stali at geology.wisc.edu>  wrote:
>> I am using solid elements and here's what I get for the smallest problem with -pc_type svd -pc_svd_monitor
>>     SVD: condition number 4.184203727999e+13, 0 of 2187 singular values are (nearly) zero
>>     SVD: smallest singular values: 9.975131304660e-01 9.979349868120e-01 9.986761984005e-01 9.993267254639e-01 9.996135846381e-01
>>     SVD: largest singular values : 3.933977100529e+13 3.933977100529e+13 4.162059658259e+13 4.162059658259e+13 4.173798159224e+13
> This is a very large condition number for a "vanilla 3D linear elasticity problem" [there are no rotations].
> The 5 lowest values are very similar … are you sure you have sufficient BCs?
>> Tabrez
>> On 08/31/2013 03:06 PM, Umut Tabak wrote:
>>> On 08/31/2013 06:25 PM, Tabrez Ali wrote:
>>>> Hello
>>>> So I used PCSetCoordinates and now GAMG seems to work really well in that the number of iterations are relatively constant. Here are the number of iterations on 4 cores
>>>> DOF       ASM    GAMG
>>>> 2187       15     22
>>>> 14739      26     22
>>>> 107811     51     29
>>> Hi,
>>> Just curious, what is the approximate condition number of this matrix, is that modelled with domain elements like solids or thin elements like shells?
>>> Best,
>>> U.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: viz.gif
Type: image/gif
Size: 146105 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20130901/8c69ed73/attachment-0001.gif>

More information about the petsc-users mailing list