ALE module run interrupted (version 3.3.2)

[vshinde]

Dear CS team,

Following the 'restart' problem with ALE module in version 3.0.4 of the code, I switched to latest version 3.3.2. The problem I encountered is, I could run ALE module with zero forces on the structure as defined below,
=============code===============
do istr = 1, nbstru
if(ntcabs.le.110000)then
forstr(1,istr) = 0.d0
forstr(2,istr) = 0.d0
forstr(3,istr) = 0.d0
else
forstr(1,istr) = 0.d0
! forstr(2,istr) = 0.d0
forstr(3,istr) = 0.d0
endif
enddo
===============================
but the run got interrupted at ntcabs=110001. The problem possibly came through fluid force 'fluid_fy' or mesh movement.

In addition, the listing file had a warning,

==========================listing===========================================
@
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
@
@ @@ ATTENTION : A LA LECTURE DU FICHIER SUITE
@ ========= AUXILIAIRE
@
@ Le nombre de structures maximal NSTRMX supporte par le
@ format d'ecriture du fichier suite est
@ NFMTST = 99
@ On a ici un nombre de structures maximal superieur
@ NSTRMX = 200
@ Si le nombre de structures effectif est superieur,
@ elles ne seront pas relues.
@
@ Le calcul sera execute.
@
@ Voir le sous-programme lecamx.
@
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
@

Fin de la lecture des donnees des structures ALE
Fin de la lecture
==========================================================================

Please find the listing.txt file attached.


Many thanks,
Vilas

[Yvan Fournier]

Hello,

The message relative to restart file format limitations is not important as long as you do not have more than 99 structures (the same warning probably exists in older versions of the code).

Do you have any other error message (error_n*), or an error (possibly related to bounds checking) in the terminal (or batch log) from which you ran the code ? It seems in your code, starting with the iteration which fails, you do not define one of the force values.

If you change the time step at which the force component 2 is not defined (for example doing it much earlier), does the code crash earlier ?

Otherwise, for smaller changes than switching to 3.3.2 (which is not a "long-term", fully validated version), I released 3.0.5 the same day as 3.3.2, and it also includes the fix for ALE restarts.

Regards,

Yvan

[vshinde]

Dear Yvan,

Thanks for the concern.

I ran another simulation and I got this below error message
============tail of listing or error file==============================
SIGSEGV signal (forbidden memory area access) intercepted!

Call stack:
1: 0x2add388321e0 <+0x321e0> (libc.so.6)
2: 0x2add37884933 <__intel_new_memcpy+0x17c3> (libirc.so)
End of stack
============================================================

Well, now I have started using v3.0.5 of the code. It works fine. I couldn't trace anything for v3.3.2 except some doubts about the name changes of mesh viscosity (which you have already clarified in another post) and force on a structure (bfprp_for).

Thanks again for the prompt reply.

Vilas