[FLASH-USERS] Checkerboard of Magnetic Monopoles

[Seyit Hocuk]

On 05/09/2012 01:28 AM, Aaron Froese wrote:
> Hello everyone,
>
> Regarding the lattice of magnetic monopoles that develops when adding the Cooling unit to a USM MHD simulation, I have induced the behaviour in multiple problems using both Flash3.3 and Flash4b.  Because the location of the lattice strongly correlates to the quadrant/octant farthest from the origin in each CPU's domain, I am concerned that it is specific to my computing hardware.  Has anyone been able to reproduce this behaviour?  What is the significance of the 8x8 squares or 8x8x8 cubes that appear?  Is that a fundamental grouping in the grid?
>
> Aaron

Hi Aaron,

I once had a problem with compiler optimization. When using "-O2" I was 
getting weird values for parent blocks. These were also effecting the 
leaf blocks eventually. Changing to "-O3" helped. I was curious if your 
problem was similar. I don't know if it were 8x8(x8) squares (cubes), 
but parents are a multiple of 2 everytime, and one block is 8x8(x8).

Another thing I have, which might be related is that I get fluctuating 
magnetic field variables (magx,magy,magz,magp) around the places where a 
high density is moving through, but divB is very low 10^-34. Could it be 
related?

Seyit


-------------- next part --------------
A non-text attachment was scrubbed...
Name: magp.png
Type: image/png
Size: 686121 bytes
Desc: not available
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20120509/02cb64e6/attachment-0001.png>