View Issue Details Jump to Notes ] Issue History ] Print ]
IDProjectCategoryView StatusDate SubmittedLast Update
0000250OpenFOAM[All Projects] Bugpublic2011-07-14 10:422011-07-15 17:39
Assigned Tomattijs 
StatusclosedResolutionnot fixable 
PlatformLinuxOSRHELOS Version5.6
Product Version2.0.x 
Target VersionFixed in Version 
Summary0000250: decomposePar does work with ptscotch or scotch
(1)When run decomposePar ontutorials/mesh/snappyHexMesh/flange with ptscotch method, there is en error as followings:

Time = 0
Create mesh

Calculating distribution of cells
Selecting decompositionMethod ptscotch
*** The MPI_comm_size() function was called before MPI_INIT was invoked.
*** This is disallowed by the MPI standard.
*** Your MPI job will now abort.
[f-2.firmflow:6683] Abort before MPI_INIT completed successfully; not able to guarantee that all other processes were killed!

(2)It seems to me that decomposePar did not been liked with scotchDecomp or metisDecomp libs...

Thanks in advance.
TagsNo tags attached.
Attached Files

- Relationships

-  Notes
mattijs (manager)
2011-07-14 11:43

Dear Guanghao,

the ptscotch method can only be run in parallel since it needs a valid mpi environment. This is what it complains about - mpi_init has not been called.

To use decomposePar you'll have to change the method to e.g. scotch or hierarchical.

As an additional constraint snappyHexMesh does not know about scotch, only ptscotch since these two libraries cannot co-exist in a single executable (this is a known limitation of scotch). So the advice is to run snappyHexMesh with the 'hierarchical' decomposition method and switch to scotch before doing decomposePar.

See e.g. the turbineSiting tutorial on how to run snappyHexMesh in parallel as well.

- Issue History
Date Modified Username Field Change
2011-07-14 10:42 guanghao New Issue
2011-07-14 11:43 mattijs Note Added: 0000552
2011-07-15 17:39 mattijs Status new => closed
2011-07-15 17:39 mattijs Assigned To => mattijs
2011-07-15 17:39 mattijs Resolution open => not fixable