[manish1308]

Hey,
I am running a simulation using parallel computing and I am using scheme commands. The issue is that I am not able to run the simulation and I think the issue is the scheme commands are running on the host node rather than the compute nodes. Are there any commands which will run the scheme commands on the compute nodes.
For eg In UDF you can use #if !RP_HOST. Something like that for scheme?

Thank you ,
Manish

[AlexanderZ]

scheme is executed on CORTEX, with is even above host node.
there is no way to execute scheme on computational nodes, as far as I know.

show yoour code, Ill try to give you advice.
sometimes there is a problem with scheme in parallel consideration

[manish1308]

Hi Alexander,
Thank you for your reply. Here are the scheme commands I am using-
(do ((x 1 (+ x 1))) ((> x 2))
(rename-file (format #f "Input_~a.c" x) "Input.c" )
(ti-menu-load-string (format #f "/file/read-case "file-name.cas.gz""))
(ti-menu-load-string (format #f "/file/read-data "file-name.dat.gz""))
(ti-menu-load-string(format #f "solve initialize initialize-flow it 50"))
(ti-menu-load-string (format #f "wc file-name-Test-~a.cas.gz" x ))
(ti-menu-load-string (format #f "wd file-name_Test-~a.dat.gz" x ))
(rename-file "Input.c" (format #f "Input_~a.c" x )))

And the input.c file is read in a udf to extract values for the further calculations.
The .c file for the udf is using #if !RP_HOST to make sure it runs on the compute node.

[AlexanderZ]

you don't read .c file in the code above.

do you interpret it somewhere else? I recommend you to compile your code( .c file)
so you can load precompiled library from scheme

[manish1308]

Hello,
Thank you for the reply. Actually the Input_1.c and Input_2.c are just files in which different numbers are stored. It was .csv file I just renamed it to .c file because the .csv file is not getting copied on the compute nodes.
So I wont be able to compile that, I mean it does not have any commands like #include "udf.h" . It just has a row of numbers. This file is read in another .c file where it picks these values and assigns them in udf. That file is compiled.

[AlexanderZ]

you should organize reading from file in UDF suitable for parallel computation
are you sure it has been done?

once again, you don't read files in scheme, you just rename them

[manish1308]

Yes, I am reading from a udf which is suitable for parallel computation.

These are the parts of the code which read the file.


#include "udf.h"
#if RP_DOUBLE
# define MY_INPUTS "%lg, %lg, %lg\n"
#else
# define MY_INPUTS "%g, %g, %g\n"
#endif

real QdotZ1;
real QdotZ2;
real QdotZ3;
real A;
real B;
real C;

DEFINE_EXECUTE_ON_LOADING(test,lib2)
{
#if !RP_HOST
int i=0;
FILE* cFile;

cFile = fopen("Input.c","r");

for (i = 0; i<1; i++)
{
fscanf(cFile,MY_INPUTS,&A, &B,&C);
QdotZ1 = A;
QdotZ2 = B;
QdotZ3= C;

}
fclose(cFile);
#endif
}

[AlexanderZ]

you are using #if !RP_HOST macro, which means, you are reading file on nodes

usually you want to read file on Host, not on nodes. But transfer data from host to nodes after reading is finished

to transfer data you need host_to_node_real macro

this could be the problem
for more information read Ansys FLuent Customization manual -> Chapter 7: Parallel consideration

[visitor]

Hi,
I can see you are experts in TUI Scheme.

Can anyone show me an example in how to read temperature, and then loop for the next cell? Please.

Or read all the cells temperature values? I can then apply an 'if ' condition statement to take a specific action depending on cell value.

Thanks.