[M_Tidswell]

Dear All

Has anyone managed to get the Ansys turbulent flame speed modification UDF to run in parallel? If so could anyone point me in the right direction.

As I understand it the modification wants to be shared by all of the computing nodes and the head node but there is no complex face/cell looping which seems to be the focus of most of the threads I can find.

I have managed to get my modification of the given example to run in serial but I have approx 5 mill elements so I will be old and grey if I have to run all the cases in serial. The latest attempt can be seen below but I still get the error:

999999 (..\src\mpsystem.c@1172): mpt_read: failed: errno = 10054

999999: mpt_read: error: read failed trying to read 4 bytes: No such file or directory
MPI Application rank 0 exited before MPI_Finalize() with status -1073741819

Latest attempt is below:

#include "udf.h"
#include "sg_pdf.h" /* not included in udf.h so must include here */
#include "turb.h"
#include "turb_ke.h"
#include "mpi.h"

DEFINE_TURB_PREMIX_SOURCE(turb_flame_src,c,t,turb_ flame_speed,source)
{
real up = TRB_VEL_SCAL(c,t);
real lx = TRB_LEN_SCAL(c,t);
real ut,ul,sct, kar, dab, leff,re_t, a, L,lf,grad_c,rho_u,Xl,DV[ND_ND];

ul = C_LAM_FLAME_SPEED(c,t);
Calculate_unburnt_rho_and_Xl(t,&rho_u,&Xl);

if(NNULLP(THREAD_STORAGE(t,SV_PREMIXC_G)))
{
NV_V(DV, =, C_STORAGE_R_NV(c,t,SV_PREMIXC_G));
grad_c = sqrt(NV_DOT(DV,DV));
}

functioning equations ..........
C_UDMI(c,t,0)=grad_c;
C_UDMI(c,t,1)= ut;
C_UDMI(c,t,2)=rho_u*ut*grad_c;
C_UDMI(c,t,3)=C_PREMIXC(c,t);
*turb_flame_speed = ut;
*source = rho_u*ut*grad_c;
}

Please forgive the ..... sections, there is merely a series of relatively standard equations which are not really mine to give out I have had a play and neither of these are the issue.

Any help would be much appreciated

Mark

[M_Tidswell]

Dear All

I thought I would post the answer to my misinterpretation as other people may do the same thing - whilst I am sharing the modification across all of the nodes it is the host node that communicates this as part of the premix equation so it is merely the host(head) node that requires the modification.

If instead of #If PARALLEL

you use If !RP_HOST


it all works rather nicely,

Thanks to http://www.cfd-online.com/Forums/flu...f-problem.html for the nudge

Mark