Learning Forum

[Ashwini Kumar]

Logging this is a tip!

 

One potential reason could be that you are trying to run a case file saved in a newer version than it is being run. In my case, I saved the case file in 25R1 and was trying to run in 24R2.

 

 

and saving it may show this error in CMD window:

 

[roberto.scipioni]

Hi there,

 

I keep getting this error

>  /solve/dual-time-iterate 1e-2 

ERROR: Invalid integer or integer not within limits [0 to #f].
Error object: 0.01

[Hit ENTER key to proceed...]
300
Maximum Number of iterations per time step [10]  

ERROR: chip-exec: function "none" not found.

ERROR: chip-exec: function "none" not found.

====================================

Anything I can do to debug this?

Details:

Ansys v222 and v242

Operating System Linux Almalinux

Intel or OpenMPI

 

Roberto Scipioni

Red oak Consulting UK

[Rob]

Put the /solve/dual-time-iterate command into an interative session and see what Fluent asks for. 

Re the chip-exec it's also associated with UDFs so check you've not got something loaded that's not necessary. 

[roberto.scipioni]

Is there documentation on using UDF with Fluent?

The thing is : I am only the sys admin so not really an expert wit Ansys Fluent :-)

Any suggestion?

R.

 

[Rob]

 

Learn? 😉  

More usefully. If you’re on Windows if you put the .c file into the folder with Fluent when you Compile (that’s covered in the manual) use the Built in Compiler. We’ve bundled the CLANG compiler with Fluent for the last few years, this negates the need to mess with Visual Studio. If you have Linux do as for Windows but no need to use the built in compiler as Linux is a proper OS. 

The Public form of the Documentation is here https://ansyshelp.ansys.com/public/account/secured?returnurl=/Views/Secured/prod_page.html?pn=Fluent&pid=Fluent&lang=en   and you want the Customisation Manual. 

The problem you’ve posted above is actually the journal isn’t correct, the argument following /solve/dual-time-iterate is an integer number of time steps, and not the time step size. Your users should be providing valid journal scripts, although I’d tend to set everything via the GUI and the batch script should just load case & data, run the job, write case & data and close. 

 

[roberto.scipioni]

We are using command line interface on an Linux HPC so it all done with text files that are edited, run ...etc

R.

[Rob]

OK, and has the .c file been compiled previously on Windows or Linux? You may need to compile before running if that's case as Win11 libraries are different to Linux: that may explain the chip-exec. One trick is to open Fluent via the GUI in the working folder and compile, you don't need a case file to do this. Then close Fluent and run the corrected journal. 

[roberto.scipioni]

OK, I had the user permission to run some test in her directory...will try and let you know...for the time being thanks :-)

 

[roberto.scipioni]

I can see there is a hierarchil file that is used .h5 is that what you are refering to?

 

[Rob]

The files they should provide are:

something.cas.h5    That's the geometry & settings 

something.dat.h5    That's the data, this can also be created by initialising the model from the case file so is optional. 

whatever.c       The UDF file that's causing you the problems

whatever.jou     The journal file that may or may not make everything work. 

 

On initialisation and/or completion of the run I'd expect additional files:

libudf     Folder that's created when you compile the .c file

something_123.cas.h5 and something_123.dat.h5   that's the output of the model. You need both files as the data is just the numbers and uses the case to make sense. 

[roberto.scipioni]

[roberto.scipioni]

Is the command:

fluent 3ddp -t40 -gu -driver null -cnf=$HOSTS -pinfiniband -cflush -i inputFile.in

valid?

 

[roberto.scipioni]

So I tested it using:

fluent 3ddp -t120  -gu -driver null -cnf=$HOSTS -pinfiniband -cflush -i  inputFile.in 

in the output we get:

...

...

>  /solve/dual-time-iterate 1 300  

ERROR: chip-exec: function "none" not found.  

ERROR: chip-exec: function "none" not found

....

....

================

How do I run 

/solve/dual-time-iterate 

interactively?

 

Many thanks

Kind regards

 

Roberto

[roberto.scipioni]

Ok got the interactive session :-)

Running it I get:

...

...

Error: An error or interrupt occurred in the previous operation.
       Details have been printed to the text console transcript.
Error Object: #f

...

...

 

=====

Any idea on how to further troubleshoot this?

Many thanks

 

Roberto

 

 

[Rob]

I assume you've got inifiband mpi? If you've got an interactive session, what did you set in parallel, and was there anything useful in the transcript (.trn file). 

The chip exec means Fluent's looking for something UDF related "somewhere". 

 

On launch I see this for interconnect

MPI Option Selected: intel

Selected system interconnect: default

 

You might need someone who can use Fluent to troubleshoot the Fluent part. Under Solution tab, and possibly under Run Calculation (my screen resolution isn't good) look for "List Modified Settings". That should tell you everything that's been changed. You're looking for a UDF (libudf) or a blank line. 

[roberto.scipioni]

Hi,

Thanks.

1) As mentioned. As we are in an HPC environment (Linux) we do not use GUIs

2)I can use fluent as part of ansys (and I did) but I cannot use the GUI.

3)I have to confess I am not sure I fully understand this libudf thing! Is it a system library or a user compiled library (very weird in a unix environment!!!)

   If it is a user compiled library do I need to retrieve the original .c file or?

 

Kind regards

Roberto

[Rob]

Libudf is the compiled library, and you'll need either libudf compiled in parallel on the cluster OS or the .c file to compile once you've launched Fluent. If you compile from Fluent before reading the case it may resolve the chip-exec as the library will be available to load. 

Is there a way to attach a viz node the cluster? 

[roberto.scipioni]

Thanks.

Is libudf part of the Ansys/fluent software or is it a separate thing?

this is what I found in the user space. Does it need to be compiled?

Should not this be included in the standard fluent directory?

 

Please advice

Kind regards

 

Roberto

 

[Rob]

The libudf folder is created by compiling a suitable .c file so that Fluent can work with the library. It's read by Fluent as part of the model set up. 

[roberto.scipioni]

Something like this?

?

R.

 

 

[Rob]

Yes. You'll need to compile some or all of the .c files. I'm fairly sure they can all be compiled at once into the one libudf, but you also need to check with the user as to what they set up. 

[roberto.scipioni]

Another question:

Where do you find these .c files does the user have to write them or?

R.

[Rob]

The user writes them. There have been a few Ansys provided modules but they're typically provided as "unsupported" in that the user is responsible for any problems. There are examples which (usually) work in the manual.

UDFs are a more advanced part of using Fluent and are used to add functions that aren't available in the core code: that could be as simple as a boundary profile, or as complex as a complete model for something. Hence the difference in support, a profile might be 10-15 lines but a full model could be 3-4000 lines of code. 

[roberto.scipioni]

so the user should be able to create and/or use them then !

R.

[roberto.scipioni]

How does fluent know that he "needs" to use these udfs?

A tag at the command line or?

[Rob]

They're added into the case file, in this case a cell zone effect. When I write & later read this case it'll look for the libudf and on finding it be happy. 

If that libudf isn't present you may get a chip exec error, or something else more terminal depending on what the UDF is for. 

[roberto.scipioni]

Thanks.

So how does the command:

/solve/dual-time-iterate 

know that has to use these UDFs ?

Is it not /solve/dual-time-iterate an internal command in fluent? 

Thanks

Roberto

[Rob]

It doesn't. The /solve/dual-time-iterate command triggers the Fluent solver to execute some iterations using the settings in the case file. That case file does know about the UDF along with the mesh, boundary settings etc. 

[roberto.scipioni]

So either the user changes the case file or she has to set the proper .c files compiled the library etc. One last question, in a Linux environment would it suffice to specify LIBRARY_PATH to have fluent use/load the library or we need to set something somewhere else?

Thanks

Regards

Roberto

[Rob]

I'd just make the user put all of the files in the working folder and make sure they use local paths in the case file. Trying to pull data over the network just opens up potential for data access hiccups. Similarly, they need to write files to the local disc to avoid network latency problems. That's currently rare, but we have software-hardware speed cycles where data writing is faster than the system can handle every now and again: ie when people run older switches/RAM on clusters and we've just updated the MPI functions. 

[roberto.scipioni]

Hi there,

I have installed separately LSdyna 16.1.1 and with both ansys242 and 252 when running it it says

I am not sure I understand this error!

Can the licence file only work for specific versions of packages?

If so what is the solution?

 

Many thanks

Roberto

[Rob]

You'll need to post into a new thread for this - I don't know anything about using Dyna. 

[roberto.scipioni]

OK thanks

 

[roberto.scipioni]

Hi,

 

Going back to The Fluent error. The user confirmed that she took the simulation from a friend. She copied input etc but the case file is DIFFERENT could it be that in changing the case file

the simulation requires the libudf library?

I converted the .h5 file into a text file. Is it written anywhere the need to use the libudf.so ?

Many thanks

Kind regards

 

Roberto

[Rob]

What exactly did she copy? If she took the setup FILE into a new mesh then that setup file might contain the UDF calls, but lacking a libudf things get interesting. 

You don't need the libudf as such, it's the macro identity from the libudf. The problem being you're possibly looking for an absence of a setting..... 

[roberto.scipioni]

...I asked the user to run exactly the same simulation of the other person who did not get the error. We will see what happens.

R.

[sc3618]

Hi, I am Sylvia the user that got the original error. I need to figure out how to use the libudf with fluent. Can you please help me?

[Rob]

OK. Have you got the .c file?  If so, it needs to be compiled for the OS and parallel for the cluster AND the libudf needs to be in the working directory on the cluster. 

[sc3618]

Hi Rob, I have recently started working with it but sure what it means. Could you please provide me some sample files ?