The ABINIT Users Mailing List ( CLOSED )

[Michel Cote <Michel.Cote@umontreal.ca>]

Hello Alex,

In our cases, we found that we need to increase the MPI_GROUP_MAX to very high values. In fact, it depends on the number of k-points. The work around for now is to increase MP_GROUP_MAX to twice the number of k-points and split the dataset in different runs using the jdtset input parameter.

We have figured out how to solve the problem but we still need to program it in abinit. Hopefully it should be done by the end of the month.

Michel

Le 07-08-02 à 08:44, a.cote@ucl.ac.uk a écrit :

I am very interested in the solution of this problem; as it stands, it is impossible for me to use abinit with MPI on an SGI ALTIX machine... abinis works fine, but abinip gives the following error: "MPI has run out of internal group entries. Please set the environment variable MPI_GROUP_MAX for additional space." 

I was quite relieved to see that I wasn't the only one with the problem - I have tried increasing MPI_GROUP_MAX from 32 (the default) up to 1024, and even though the calculation goes further, it eventually always crashes.

Is there a solution to free the MPI groups?

Thank you,

Alex

============================

Dr. Alex Côté

Department of Earth Sciences

University College London

Gower Street

London WC1E 6BT

Tel: +44-(0)-20-7679-2425

Email: a.cote@ucl.ac.uk

============================

_____________________________________

Michel Cote

Departement de physique

Universite de Montreal

emai: Michel.Cote@umontreal.ca

web: http://www.phys.umontreal.ca/~michel_cote