[CIG-SHORT] Step 18 with two loads

[Brad Aagaard]

Jeanne,

When I looked at your input files, I realized you have one load, but are 
changing the spatial variation of where the load is.

You are not getting the full error message. I think this is an MPI 
related issue on Macs that we don't have control over. It may eventually 
go away as we switch to updated versions. When I run your step18.cfg 
with the v1.9.0 binary on my Linux machine, I get:

RuntimeError: Could not find values at ( 211.325 1788.68 0) for traction 
boundary condition face_zpos using spatial database Neumann BC on +z.

The problem is that we use a very simple interpolation algorithm for 
areal interpolation. It works well for relatively evenly spaced data and 
breaks down for very irregularly spaced data. Your additional point 
resulted in some unevenly spaced points. The solution is to fill in more 
points on the surface to even out the spacing.

For example, add points along the center of the edges:
  3.0  0.0
  0.0  3.0
-3.0  0.0
  0.0 -3.0

For more points, add
  1.5  1.5
  1.5 -1.5
-1.5 -1.5
-1.5  1.5

The points don't have to be perfectly evenly spaced and you can 
transition the spacing between points, but you need to do it gradually.

We should really add a third interpolation algorithm that uses radial 
basis functions for cases when users what to use very irregularly spaced 
points. I will add this to the long-term PyLith TODO list.

Regards,
Brad


On 01/09/2014 12:42 PM, Sauber-Rosenber, Jeanne M. (GSFC-6980) wrote:
> Brad,
>
>    Sorry I didn't send these earlier:
>
> Case 1:  Example Step 18 except with two surface loads (attached spatialdb for input)
> "db_initial.iohandler.filename = spatialdb/tractions_axial_pressure.spatialdb.txt" ,
> see attached).
>
>> gs698-sauberair:hex8 jsauberr$ pylith step18.cfg > step18L.out
>
> Error (see full output in step18L.out attached as well):
>
> mpinemesis(81253) malloc: *** error for object 0x104d09934: pointer being freed was not allocated
> *** set a breakpoint in malloc_error_break to debug
> [0]0:Return code = 0, signaled with Abort trap: 6
>
> Case 2:  Step 18 BUT with query = nearest, same input file as Case #1
>
> Output:  No errors as Case 1, see step18N.out
>
> Thanks,
> Jeanne
>
>
>
> _______________________________________________
> CIG-SHORT mailing list
> CIG-SHORT at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>