[uconcorde]

Hi Everyone,

I am using udf "DEFINE_PROFILE" to get a boundary layer velocity profile for the inlet of my computational domain.

My udf is very simple and works perfect in Serial Mode.

Then, I followed the examples in UDF manual and tried to parallelize my udf. It seems that the face_loop in the udf is only able to loop in one partition (see the lower section of the parallel plot, which gives correct values), and values of the upper section are not assigned correctly.

Sorry, I don't know why I cannot attach these images, please click them
serial mode profile: https://drive.google.com/file/d/1kQb...ew?usp=sharing
parallel mode profile: https://drive.google.com/file/d/1gO0...ew?usp=sharing

Here is my udf:

DEFINE_PROFILE(inlet_u_profile,t,i)
{
real xf[ND_ND];
face_t f;
float blasiusU[] = { ...blasius values... }; /* store blasius profile data */
#if !RP_HOST
begin_f_loop(f,t)
{
F_PROFILE(f,t,i) = blasiusU[f];
}
end_f_loop(f,t)
#endif
}

Does anybody know how to continue looping the faces on the adjacent partition in the domain?

Thanks in advance.
Dv

[pakk]

float blasiusU[] = { ...blasius values... }; /* store blasius profile data */

This is unusual for me. For this to work, you need to store the values of the faces in the same order as Fluent wants to read them. This order is never specified, as far as I know. If it worked for you in the serial case, I guess that you were just 'lucky', as Fluent took the same order as you guessed.

But in the parallel case, Fluent will (apparently) take a different order. So your approach does not work.

The normal approach for this: describe the profile as a function of x (or whatever coordinate that varies).

float blasiusU(float x) {
 if (x<0.03) {   /*just an example function */
  return(0.2-x);
 } else if x<0.6 {
  return(0.173-0.1*x);
 } else {
  return(0.113);
 }
}

DEFINE_PROFILE(inlet_u_profile,t,i)
{
#if !RP_HOST
 real xf[ND_ND];
 face_t f;
 begin_f_loop(f,t) {
  F_CENTROID(xf,f,t);
  F_PROFILE(f,t,i) = blasiusU(xf[0]);
 }
 end_f_loop(f,t)
#endif
}