Hi, there has been some problem when I try to load the compiled udf library in parallel computing.It says

library "libudf/lnx86/2ddp_host/libudf.so" opened Primitive Error at Node 0: open_udf_library:No such file or directory Primitive Error at Node 1pen_udf_library:No such file or directory Error: Compute node encountered error Error Object:#f

However, everything runs ok when I run it on just one computer, not parallel, so there is no problem with the udf itself, but some problem of whether the library can be seen by all the computers. Hope there'll be someone who can give me some idea. Thanks in advance. Lucy

The library must be accessible to all computers in the cluster.

See UDF manual:

5.7

5.2


" If you are running FLUENT on a network of Windows machines , you may need to type the file's complete path in the Source File Name field in the Interpreted UDFs panel, instead of using the browser option. For example, to compile udfexample.c from a shared working directory named mywork, you would enter the following in the Source File Name field:

\\<fileserver>\mywork\udfexample.c

figure

Here, you replace <fileserver> with the name of the computer on which your working directory ( mywork) and source file ( udfexample.c) are located."

Hi, My earlier post to a similar query answers your problem.

---------------------------------- The possibilities are: 1. Are you usung UDF? Yes then recompile your UDF on all the nodes. I assume you are using SGE for your work/job submission. In that case open a terminal and type in Fluent &. After the console pops up click on File, run. In that console check parallel and then select number of processors. After that compile your udfs and then exit. You are ready to submit your script now to SGE.

2. You can follow the same procedure to read your case and data. The only difference would be not to follow the step of submission to SGE.

3. Please remember this error is might not be due to network communications but only due to the fact that your libudf libraries are not been acessible by slave nodes.

I would assume in the discussion that you clsuter facility has all the nodes shared and you have read and write acess on them!

Last but not the least: If you are having customised data writing functions over the parallel environment you will have to modify your UDF's slightly. This has been explained in the UDF manual. Let me know how it goes and you can also reach me at kp@dal.ca. Good luck! KP

[Kanarya]

Hi

I have problem to run in parallel this code?It is simple C_UDMI which is calculated before:
DEFINE_MASS_TRANSFER(water1,c,t,from_index, from_species_index, to_index, to_species_index)
{
#if !RP_HOST

return(C_UDMI(c,t,10));

#endif

}

in serial it is ok but parallel giving me following error:
warning C4716: 'water1' : must return a value.

is there anyone had expericed before?

Thanks in advance!!!