Bug in compute wall/gran/local

Submitted by tapsab on Tue, 09/11/2012 - 20:10

There seems to be a bug in "compute wall/gran/local" when multiple meshes exist in a simulation. The mesh id's reported by the compute are erroneous. I attach an example to illustrate the same. The simulation has two particles tightly bound between three walls, bottom, left and right. As per the order in the command "wall/gran/hertz/history" the mesh id's should be bottom->0, left->1, right->2. The bottom wall is touching one particle and the left and right walls are touching the both the particles. However when you look at the dump output (stored in post/dump.wc) the bottom is reported to be in contact with both particles. The error here is that for one of the contacts, the mesh id should be 2, corresponding to the right wall, instead the reported mesh id is 0, corresponding to the bottom.

A couple of other notes. There also seems to be a bug in the initialization of the output obtained from "compute wall/gran/local". If you run the simulation on multiple cores, the are a bunch of rows in the output with zeroes. However a few times, instead of zeroes I would get values like 300e-12. For example see the output below obtained by running the attached simulation with 4 cores. Note that even the mesh id "0" is reported instead as 6.9211e-310 or 6.91062e-310.

Thanks,
Tapan

ITEM: TIMESTEP
0
ITEM: NUMBER OF ENTRIES
20
ITEM: ENTRIES c_wc[1] c_wc[2] c_wc[3] c_wc[4] c_wc[5] c_wc[6]
6.9211e-310 6.9211e-310 1 -0 959624 2.13079e-11
0 0 1 959624 -0 2.13079e-11
1 0 1 -959624 -0 -0
2 0 0 0 0 0
0 0 0 0 0 0
6.91141e-310 6.91141e-310 1.58101e-322 1.58101e-322 2.12799e-316 2.122e-314
3.16202e-322 1.58101e-322 2.128e-316 2.122e-314 4.74303e-322 1.58101e-322
0 2.122e-314 2.122e-314 5.22573e-320 0 0
0 0 0 0 0 0
0 0 0 0 0 0
6.91062e-310 6.91062e-310 2 959624 -0 2.13079e-11
1 1 2 -959624 -0 -0
2 1 2.122e-314 6.2735e-319 0 0
0 0 0 0 0 0
0 0 0 0 0 0
6.90477e-310 6.90477e-310 1.58101e-322 1.58101e-322 1.95821e-316 2.122e-314
3.16202e-322 1.58101e-322 1.95821e-316 2.122e-314 4.74303e-322 1.58101e-322
0 2.122e-314 2.122e-314 5.20202e-320 0 0
0 0 0 0 0 0
0 0 0 0 0 0

AttachmentSize
Binary Data Bug.tar_.gz3.03 KB
ckloss's picture

ckloss | Mon, 09/17/2012 - 01:12

Hi Tapan,

thanks for reporting this!
I will issue a fix for this in the next release!

Cheers, Christoph

msbentley's picture

msbentley | Thu, 09/20/2012 - 16:23

I guess this is already clear? If not, another data point :) I've also hit the problem of odd dump/wall/local output - specifically the reporting of many more particle/wall contacts than are real, the "fake" entries all appearing as zeros (or very small numbers).

I have a system with 800 particles in contact with the wall, and the dump shows this for a single CPU. For a run with 2 CPUs, however, the dump shows ITEM: NUMBER OF ENTRIES 1600, and with 3 CPUs 2400 etc. There are indeed 2400 entries - the real 800, and the rest zeros.

Cheers, Mark