Welcome to the forums. Please post in English or French.

You are not logged in.

#1 2022-07-29 17:43:23

stephaneberger
Member
From: Strasbourg (France)
Registered: 2012-10-15
Posts: 70

CALC_MODES and MATR_DISTRIBUEE

Hi

I'm trying to run a modal analysis with code_aster_MPI.

modes = CALC_MODES(CALC_FREQ=_F(FREQ=(200.0, 6000.0, 12000.0),
                                NIVEAU_PARALLELISME='COMPLET',
                                NMAX_ITER_SHIFT=5),
                   MATR_MASS=MASSE,
                   MATR_RIGI=RIGIDITE,
                   OPTION='BANDE',
                   SOLVEUR=_F(
                              METHODE='MUMPS',
                              GESTION_MEMOIRE='AUTO',
                              ),
                   SOLVEUR_MODAL=_F(METHODE='SORENSEN'),
                   TYPE_RESU='DYNAMIQUE',
                   VERI_MODE=_F(STOP_ERREUR='NON'))

My model is quite large and the solve needs huge amount of RAM.
In the log file, this is written that I have to activate MATR_DISTRIBUEE='OUI'.
If I activate this solver option, I have an error

ERREUR A LA VERIFICATION SYNTAXIQUE - INTERRUPTION
>> JDC.py : DEBUT RAPPORT
DEBUT CR validation : fort.1
   Etape : CALC_MODES    ligne : 27    fichier : 'fort.1'
      b_bande
         SOLVEUR
            !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
            ! Mots clés inconnus : MATR_DISTRIBUEE !
            !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
         Fin SOLVEUR
      Fin b_bande
   Fin Etape : CALC_MODES
FIN CR validation :fort.1

In the documentation U4.50.01, this is written (page 18/25):

MATR_DISTRIBUEE = / ‘YES’
                  / ‘NOT’ [DEFECT
This parameter is for the moment limited to the operators MECA_STATIQUE, STAT_NON_LINE
and DYNA_NON_LINE ....

To me, this is not possible to use MATR_DISTRBUEE inside a CALC_MODE.
Does anyone has the answer ?

Thanks in advance

Stephane

Offline

#2 2022-08-04 10:11:24

RichardS
Member
From: Munich, Germany
Registered: 2010-09-28
Posts: 560
Website

Re: CALC_MODES and MATR_DISTRIBUEE

Hi Stephane,
my guess is that the message in the log file is very generic and doesn't change
depending on the context from where the MUMPS solver is used (e.g. the analysis type).
Also based on the documentation and your test MATR_DISTRIBUEE simply isn't available in CALC_MODES.

I guess you need to find a different approach for reducing the memory demand of the Egenmode calculations (which indeed is very high in Code-Aster based on our experiences as well).

Best,
Richard


Richard Szoeke-Schuller
Product Management
www.simscale.com
We are hiring! https://www.simscale.com/jobs/

Offline