[FLASH-USERS] MPI issue

[Christos Stavrou]

PS: The OS is indeed Ubuntu 21.04 and the MPI used is the apt-installed MPICH implementation.

From: Christos Stavrou
Sent: 03 July 2021 21:40
To: Tony Curtis <anthony.curtis at stonybrook.edu>
Cc: flash-users at flash.uchicago.edu
Subject: RE: [FLASH-USERS] MPI issue

Hi Tony,

Many thanks for your help.

The setup is a single (multicore AMD) machine and I run it locally.

The series of steps, for the Sedov example, is:
./setup Sedov -auto                      (I have tried with “+noio” as well)
cd object
make
mpirun -np 10 ./flash4

Everything appears to go as intended up to the point which I shared in my previous email, at which point the code stops.

Kind regards,

Christos
From: Tony Curtis <anthony.curtis at stonybrook.edu<mailto:anthony.curtis at stonybrook.edu>>
Sent: 02 July 2021 23:01
To: Christos Stavrou <cstavrou at magdrivespace.com<mailto:cstavrou at magdrivespace.com>>
Cc: flash-users at flash.uchicago.edu<mailto:flash-users at flash.uchicago.edu>
Subject: Re: [FLASH-USERS] MPI issue



On Jul 2, 2021, at 12:06 PM, Christos Stavrou <cstavrou at magdrivespace.com<mailto:cstavrou at magdrivespace.com>> wrote:

Hi all,

Thank you to Tony for his help earlier; this next issue may or may not be related with the one I had earlier (compiling with GCC10).

At the moment, I can successfully run setup to create the object directory, enter the object directory and run make successfully, albeit with a fair amount of warnings (e.g. "Warning: Type mismatch between actual argument at (1) and actual argument at (2) (REAL(8)/INTEGER(4)" from a call to MPI_Bcast).

When I subsequently try to run the code, it goes through some of the initialisation, but stops and gives me an MPI-related error message. The example below is for the case of:

./setup Sedov -auto
cd object
make

The error message:

Initial dt verified
 *** An error occurred in MPI_Comm_create_keyval
 *** reported by process [222691329,0]
 *** on communicator MPI_COMM_WORLD
 *** MPI_ERR_ARG: invalid argument of some other kind
 *** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
 ***    and potentially your MPI job)


How did you run the program?  Is this a cluster or a single machine?

Looks like you’re using an ubuntu-installed mpich, so that should be good to go.

Tony

CAUTION - THIS EMAIL IS FROM AN EXTERNAL SENDER
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20210703/230b7ebf/attachment-0001.htm>