N-S solver can't run in multi-processor in Code-Saturne3.0.5

[yjwudi]

Hi,
I met a problem with Code-Saturne3.0.5.(We using this version, because our teacher has added some functions in Code-Saturne3.0.5)
I installed hdf5, cgns, and metis-5.1.0(with shared=1 when compiling metis). Then I installed Code-Saturne3.0.5 which is downloaded from this website. Then, I installed it.
After installation(without error), I run a 3 dimension grid in one processor, it runs well. However, when I run it in 2 or more processor, it stoped! I checked the listing file under the RESU folder it stoped in the output:
Criterion 4: Cells Volume Ratio:
Number of bad cells detected: 0 --> 0 %

Criterion 5: Guilt by Association:
Number of bad cells detected: 0 --> 0 %

Computing geometric quantities (0.785 s)
And the process cs_solver is still running, but there is no more output. Where is the problem？ In the installation or somewhere else? (I have intel mpi and openmp)

[Yvan Fournier]

Hello,

Could you provide the details (except the mesh) listed in the forum recommendations ?

Also, what changes did your professor add ? Do you have a modified source tree, or simply user subroutines ? The hang could be due to those...

Regards,

Yvan

[yjwudi]

Hello,

Could you provide the details (except the mesh) listed in the forum recommendations ?

Also, what changes did your professor add ? Do you have a modified source tree, or simply user subroutines ? The hang could be due to those...

Regards,

Yvan

We have found the problem. We compile cs with "./configure ... --with-openmp ...", after we remove "--with-openmp", it will run in one node with multi processors. Can you explain the reason? Maybe openmp can only be used when we run in different nodes. Also, can you tell us how to run in different nodes

[Yvan Fournier]

Hello,

There might still be bug in our OpenMP code (it is not enabled by default in version 4.0 because it was not tested enough).

If you post your case setup, maybe I can reproduce the bug and fix it.

Regards,

Yvan