[chubb87]

Hi,

my case also contains an ACMI and I had the same error while decomposing, which I can avoid by using your "hack", thanks Bruno for the effort.

Yet when I start a solver in parallel with that case, I get this MPI-error:

*** An error occurred in MPI_Recv
 *** reported by process [715390977,1]
 *** on communicator MPI_COMM_WORLD
 *** MPI_ERR_TRUNCATE: message truncated
 *** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
 ***    and potentially your MPI job)

I get this error with a "hacked" OF-2.3.1 version and also with OF-2.4.x, which seems to be able to decompose without the hack, yet yields the same MPI-error for the solver.
Does anyone else have this problem?

BTW I tried the same case with an AMI and that works in parallel. The reason I would lik to use the ACMI setup is the bad scaling behaviour of the AMI setup.

[wyldckat]

Greetings chubb87,

Originally Posted by chubb87

I get this MPI-error:

*** An error occurred in MPI_Recv
 *** reported by process [715390977,1]
 *** on communicator MPI_COMM_WORLD
 *** MPI_ERR_TRUNCATE: message truncated
 *** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
 ***    and potentially your MPI job)

That's a waaaaaaaaaaaaaaay too vague of an error message. Without an example case, I can't reproduce the same error in order to attempt to diagnose the problem.

Originally Posted by chubb87

BTW I tried the same case with an AMI and that works in parallel. The reason I would lik to use the ACMI setup is the bad scaling behaviour of the AMI setup.

Uhm... Saying "Wrong way, go back!" comes to mind. Because ACMI is an implementation specific for having AMI surfaces that are partially blocked: http://www.openfoam.org/version2.3.0/ami.php
If your case runs with AMI, then you probably shouldn't even try ACMI... furthermore, ACMI only works in some cases, it's not yet fully operational: http://www.openfoam.org/mantisbt/view.php?id=1450#c3913

Best regards,
Bruno

[chubb87]

Hi Bruno,

then I'll probably have to stick with the AMI setup. The ACMI setup of course has the interface at a lower radius, so that it is not needed for the whole 360° but only for about 1 % of the area, where an outlet hole exists (see schematic mesh in the picture). The rest is to be modelled as a rotating wall (blocked region), I am not sure if this is possible with the ACMI.

This reduced area of the ACMI compared to the AMI should reduce the computational demand ernormously, if it would work.

[wyldckat]

Quick answer: That image looks familiar... ah HA, found it: http://www.cfd-online.com/Forums/ope...cmi-patch.html

Here are some related threads on this topic:

• http://www.cfd-online.com/Forums/ope...interface.html
• http://www.cfd-online.com/Forums/ope...h-regions.html