open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

I'm running a case with T2D V6.0
My mesh is almost 1 million nodes.
The run stops saying :
"forrtl: severe (41): insufficient virtual memory"
I guess this doesn't help. I give the whole message below.
I'm wondering about the "MAXIMUM NUMBER OF POINTS AROUND A POINT", which is 11.
Does this number should be less than 10 ?
How can i find out the number of the node that has eleven neighbours ?

Or do you think the error comes from somewhere else ?

NOMBRE D'ELEMENTS:945441
NOMBRE REEL DE POINTS:493253
MXPTEL (BIEF) : NOMBRE MAXIMUM D'ELEMENTS VOISINS D'UN POINT : 10
NOMBRE MAXIMUM DE POINTS VOISINS D'UNPOINT : 11
SEGBOR (BIEF) : NOMBRE DE SEGMENTS DE BORD = 42003
EN COMPTANT CEUX DUS A LA DECOMPOSITION DE DOMAINE
CORRXY (BIEF) : PAS DE MODIFICATION DES COORDONNEES

MAILLAGE : MESH ALLOUE



****************************************
* FIN DE L'ALLOCATION DE LA MEMOIRE : *
****************************************

INBIEF (BIEF) : MACHINE NON VECTORIELLE (SELON VOS DONNEES)

NOMBRE DE FRONTIERES LIQUIDES : 4

CORFON (TELEMAC2D) : PAS DE MODIFICATION DU FOND

CONDITIONS PARTICULIERES LUES

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

ITERATION 0 TEMPS : 0.0000 S

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

ITERATION 1 TEMPS : 5.0000 S
LE FICHIER DES FRONTIERES LIQUIDES CONTIENT
24 LIGNES AVEC :
Q(1) Q(2) SL(3) SL(4)
FRONTIERE LIQUIDE : SL(3) = 45.4500277777778
FRONTIERE LIQUIDE : SL(4) = 45.4500277777778
FRONTIERE LIQUIDE : Q(1) = 674.947222222222
FRONTIERE LIQUIDE : Q(2) = 0.000000000000000E+000

---

ETAPE DE CONVECTION
STREAMLINE: USING PARALLEL VERSION OF CHARACTERISTICS 6.0
forrtl: severe (41): insufficient virtual memory
Image PC Routine Line Source
out11548_win.exe 0060FF8A Unknown Unknown Unknown

Stack trace terminated abnormally.
forrtl: severe (41): insufficient virtual memory
Image PC Routine Line Source
out11548_win.exe 0060FF8A Unknown Unknown Unknown

Stack trace terminated abnormally.
unable to read the cmd header on the pmi context, Error = -1
.
Error posting readv, Une connexion existante a d¹ Ûtre fermÚe par lÆh¶te distant
.(10054)
unable to read the cmd header on the pmi context, Error = -1
.
Error posting readv, Une connexion existante a d¹ Ûtre fermÚe par lÆh¶te distant
.(10054)

job aborted:
rank: node: exit code[: error message]
0: 3429-SIMUL: 41: process 0 exited without calling finalize
1: 3429-SIMUL: 41: process 1 exited without calling finalize
received suspend command for a pmi context that doesn't exist: unmatched id = 0
Duree du calcul : 9 secondes ( 0:0:9 ) (systeme=0 sec)
______________________________________________________________________________
*** RESTITUTION DES FICHIERS ***

- FICHIER DES RESULTATS : RES.res

dans restihp (regroupe pour 2 processeurs)

________________________________________________________
Execution terminee: telemac2d.bat
________________________________________________________
No compilation/linking/file errors detected.
No execution errors detected.


Thank you for any help,

Regards,

Benoît D.

Hello,

No, if it were the maximum of points around a point, there would be a clearer stop. It seems that you run out of memory when you allocate some extra local memory in module streamline (most probably in subroutine organize_chars, this could be traced by adding flags in it), so it could "just" be that you do not have enough memory on your machine. Actually I can't do 1 million points either on my old Unix HP machine, but up to 3 millions is OK on our latest Linux machines. A possible solution would be choosing other advection schemes than characteristics, but with a risk of allocating an extra matrix and have the same problem elsewhere (but another memory crash elsewhere would be a proof that you do not have enough memory).

With best regards,

Jean-Michel Hervouet