I'm having problems running Fluent in the batch scheduler (background) with a UDF that I wrote. here's the UDF:

#include "udf.h"

DEFINE_PROFILE(unsteady_velocity, thread, position)

{ face_t f;

begin_f_loop(f, thread)

{

real t = RP_Get_Real("flow-time");

F_PROFILE(f, thread, position) = 0.01789 + 0.001789*sin(20.*t);

} end_f_loop(f, thread) }

the UDF calculates unsteady velocity in time and is used in an inlet boundary condition. when I run Fluent interactively, it runs fine and is able to find the UDF. when I go for background processing, the following error comes up:

Error: chip-exec: function "unsteady_velocity" not found. Error: chip-exec: function "unsteady_velocity" not found. Error: chip-exec: function "unsteady_velocity" not found. Error: chip-exec: function "unsteady_velocity" not found. Error: chip-exec: function "unsteady_velocity" not found. Error: chip-exec: function "unsteady_velocity" not found.

the error statement comes up 6 times because (I guess)I'm using 6 processors. I'm positive, I've specified the pathname correctly.

PLEASE HELP!

Do you use the compiled UDF ? If you use it, check that you create 2d(3d)_host and 2d(3d)_node sub-directories in your UDF directory.

I'm using an interpreted UDF. will this not work while parallel processing ? I see that "udfconfig_nodeX.h" files are created in my home directory. the X corresponds to (1 - No. of processors) that I use. the files are always created in my home directory and not the directory from where I invoke the batch scheduler.

should I be using compiled UDFs ? I'm working on it right now.

thanks Park anyway ....... Prateep

This is probably a scripting problem when you load in you udf. I assume your using text commands to script fluent.

I had some similar problems selecting my udfs using the text interface. I remember I had to put something like:

udf "udf_name"

to select my udf properly - the first tells fluent you want ot use a udf for the profile and the send tells it the udf to use - you need the quotes. Now you can test your script in the gui - just use text commands, when you get things going, save the commands to a journal file and use in batch mode.

Try on a single processor first - its not likely to be a problem with parallel.

Greg

Note in the previous udf and "udf name" are on different lines, like

udf

"udf name"

Greg

thanks greg for the suggestion. I'll try to implement them.

in the meantime, I've been checking fluent's support pages. I'm uploading text files to my webpage, which help solve problems with parallel UDF usage.

http://filebox.vt.edu/users/pchatter/UDF

thanks again!

prateep

Yeah I had to get some of that stuff from Fluent too. Thanks for the links. Its a pity this stuff isn't in the manual!

Greg