problems using "accumulations (weightedAverage)" in parallel and...

Status: Beta

Brought to you by: bgschaid, hakan_nilsson, henrus, hjasak, and 2 others

#196 problems using "accumulations (weightedAverage)" in parallel and "accumulations (sum)" gives twice the expected value

Status: acknowledged

Owner: Bernhard Gschaider

Labels: None

Priority: normal

Severity: minor

Reproducibility: always

Projection: none

build:

category:

fixed_in_version:

os:

os_build:

platform:

target_version:

version:

component: other

Updated: 2014-01-24

Created: 2014-01-09

Creator: Sam

Private: No

Problems using "accumulations (weightedAverage)" in parallel and "accumulations (sum)" gives twice the expected value.
---- additional_information ----
http://www.cfd-online.com/Forums/openfoam/128281-swak4foam-problems-using-accumulations-weightedaverage-parallel.html

Discussion

Bernhard Gschaider - 2014-01-09

I'll have a look

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Bernhard Gschaider - 2014-01-19

Tried a modified case as described in http://www.cfd-online.com/Forums/openfoam/128281-swak4foam-problems-using-accumulations-weightedaverage-parallel.html#post469145 and got similar results (only differences in the last few digits) of

Expression UMean : weightedAverage=(0.133544 -0.00100838 0.00157697)
Expression UMean_SUM : sum=(400.632 -3.02513 4.73091)

on 1, 2 and 4 CPUs. So it seems that the problem was solved since the last release.

Can you confirm that the above values are what you should get for that case.

Please try the version from the Mercurial-repository

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Sam - 2014-01-20

In this case the mesh is (80 50 30) which means 1 500 cells in the yz-plane and 0.133544 x 1 500 = 200.316 which is exactly half of the summed value of 400.632 so I think that the summed value still is two times too large or am I missing something?

Regarding the problems regarding running weightedAverage in parallel, it most likely has to do with the error message I get in the end of the decomposePar log file

"decomposePar: symbol lookup error: /home/xfreds/OpenFOAM/xfreds-2.0.x/platforms/linux64GccDPOpt/lib/libswak4FoamParsers.so: undefined symbol: _ZN4Foam10sampledSet28destroywordConstructorTablesEv"

since my running log file gives me

"[1] --> FOAM FATAL ERROR:
[1] Can not construct weight field of the expected size. For sizes on the processors see above
[1]
[1]
[1] From function SampledSurfaceValueExpressionDriver::weightsNonPoint
[1] in file SampledSurfaceValueExpressionDriver.C at line 316."

Is that also something that has been changed in the Mercurial version or something else I need to check up during my compilation etc?

/Sam

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Bernhard Gschaider - 2014-01-24

I don't understand the last paragraph: you are having these problems with the released version and ask whether they are fixed in the Mercurial-version? Or you're having the problems with the Mercurial-version and ask whether they are normal?

In the second case I'd be thankful if you could provide a case to reproduce the problem. In the first version please try the Mercurial-version. It's over half a year of fixes ahead of the release

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Log in to post a comment.