CFD Online Discussion Forums

CFD Online Discussion Forums (
-   OpenFOAM Running, Solving & CFD (
-   -   Probes and multiprocessor runs (

grtabor June 12, 2007 06:54

Dear All, I'm trying to get
Dear All,

I'm trying to get my head around something here. I've been running (with a student) LES calculations on 12 processors in parallel. Used decomposePar to split the mesh giving processors 0 - 11, and have successfully run the simulation. So far, so good.

We also wanted to sample the data at specific points (9 points in all) along the geometry to evaluate energy spectra etc. Established a probleLocations dictionary in constant before doing the decomposition. For a normal (single processor) run, this generates a probes directory containing probe data in files (split up according to timestep). For the parallel run, we appear to have probes directories in 4 of the 12 processor directories; each probe directory contains files each of which appears to contain data for all 9 probe locations. However the data is different from each processor (not much, but significantly - 4th sig fig in most cases).

I'm trying to understand what is happening here? Which probes directory should we be looking at? We are (for complicated reasons) still running 1.1, so this might be a "feature" which has been resolved in later versions. Any suggestions?


mattijs June 12, 2007 13:59

Each probes/ directory will on
Each probes/ directory will only give the probes which are on the local processor.

(and if none of the probes are on the local processor it will write no probes/ directory)

In 1.4 the probes are properly collected onto the master processor.

All times are GMT -4. The time now is 10:36.