Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

Subject: Re: [OMPI users] MPI-IO puzzlement
From: Secretan Yves (Yves.Secretan_at_[hidden])
Date: 2012-05-15 09:30:19


If, by any bad luck, you use any of the following FORTRAN function


they all are stiil overflowing ( because they cast the correct result to MPI_Fint which default size if 32 bits.

Yves Secretan

Avant d'imprimer, pensez à l'environnement

-----Message d'origine-----
De : users-bounces_at_[hidden] [mailto:users-bounces_at_[hidden]] De la part de Ricardo Reis
Envoyé : 15 mai 2012 04:29
À : Open MPI Users
Objet : Re: [OMPI users] MPI-IO puzzlement

  Hi all

  The problem has been found.

  I'm trying to use MPI-IO to write the file with all processes taking part in the calculation writing their bit. Here lies the rub.

  Each process has to write a piece of DIM = 35709696 elements.

  Using 64 processes the ofsett is my_rank * dim

  and so... the ofset size, for last processes becomes.

  DBG: 60 will WriteMPI_IO. dim 35709696 offset 2142581760
  DBG: 61 will WriteMPI_IO. dim 35709696 offset -2116675840
  DBG: 62 will WriteMPI_IO. dim 35709696 offset -2080966144
  DBG: 63 will WriteMPI_IO. dim 35709696 offset -2045256448

  offset is of the type MPI_OFFSET_KIND which seems insuficient to hold the correct size for the offset.

  So... am I condemned to write my own MPI data type so I can write the
files? ideas... ?

  best regards,

  Ricardo Reis

  'Non Serviam'

  PhD/MSc Mechanical Engineering | Lic. Aerospace Engineering

  Computational Fluid Dynamics, High Performance Computing, Turbulence

  Cultural Instigator @ Rádio Zero

  contacts: gtalk: kyriusan_at_[hidden] skype: kyriusan

  Institutional Address:

  Ricardo J.N. dos Reis
  IDMEC, Instituto Superior Técnico, Technical University of Lisbon
  Av. Rovisco Pais
  1049-001 Lisboa

                       - email sent with alpine 2.00 -