[Bal21]

Hi all,

I'm trying to test parallel meshing on my computer. So I've enabled the "parallel" option and run the Java script. The mesher models that I'm using are polyhedral, prismatic and surface remesher. I'm running the case on 12 cores & am not really that impressed with the performance. The license that my university has is an educational license.
What could the reason be? What to do? Any advice?

I can also test the star case on a higher number of cores in the HPC (linux system). Anyone has tried to run the meshing java script on a linux system? I'm also worried that increasing the number of cores does nothing???!!

Thanks!!

[LuckyTran]

The parallel meshing option isn't a placebo button if that's what you are asking.

Use some system monitoring tools (e.g. task manager in Windows) to check the performance of your system. Check the cpu usage. Check your ram usage and throughput. Look for bottlenecks. If you run out of RAM for example, expect the meshing to take forever.


What kinds of timescales are we talking about here? If your mesh takes more than an hour than you probably exceeded some physical limit.

[Bal21]

Meshing has been ongoing since 12 hours.

[Bal21]

So, i was able to create a mesh in star ccm in windows*using a .java file. However, when I tried to do the same in linux*system, there was an error saying that the macro file couldn't be opened

*Any thoughts?

[LuckyTran]

Are we debugging parallel Star here or is this your first attempt at meshing anything, like ever?

If it takes more than an hour on a 12-core build then you probably ran out of RAM. Have you checked the task manager!? Or try using coarser settings.... An hour of meshing on a 12-core build is gonna grab about 100GB of RAM.

If it says the file could not be opened then it means just that. It couldn't find the file or the file couldn't be opened due to the user not having read permissions (very unlikely). If it found the file and the .java was absolute garbage inside, it would give you all sort of errors with bad declarations and such. If you give it a blank .java macro, it will do nothing.

[Bal21]

To start with, I'm a new user of star ccm and also new to the CFD world. I'm a chemical engineering student, with almost no programming background.

I thought that I can learn in this forum by discussing with experts who can communicate their "wisdom" :P

I opened the task manager and it seems that the memory usage is high. The mesh is running though, but it's taking alot of time.

[Bal21]

The error with the macro file, was as you mentioned, a simple one.
It's related to the names of the files & whatever is mentioned in the linux configuration script. This has been resolved. Thanks!!

Originally Posted by LuckyTran

Are we debugging parallel Star here or is this your first attempt at meshing anything, like ever?

If it takes more than an hour on a 12-core build then you probably ran out of RAM. Have you checked the task manager!? Or try using coarser settings.... An hour of meshing on a 12-core build is gonna grab about 100GB of RAM.

If it says the file could not be opened then it means just that. It couldn't find the file or the file couldn't be opened due to the user not having read permissions (very unlikely). If it found the file and the .java was absolute garbage inside, it would give you all sort of errors with bad declarations and such. If you give it a blank .java macro, it will do nothing.

[LuckyTran]

Originally Posted by Bal21

I opened the task manager and it seems that the memory usage is high. The mesh is running though, but it's taking alot of time.

Of course the mesher is "running." It will be running until it finishes. And it might not ever finish.

Look at your cpu usage and make sure all 12 cores are running at 100%. That means it is actually meshing. Look at your RAM and make sure it's well below 90%. If you ran out of RAM then Star will keep trying to use the page file before throwing an out of memory error, which can take a very long time. If more than a few of your cores are sitting idle then it's not doing anything. I'm going to keep echoing this statement unless you tell me your machine has 1TB of RAM of some crazy amount.

When you get Star running on your cluster, all the same applies except you'll be using not Windows Task manager.

[Bal21]

Thank you. I did monitor the performance & exactly as you suggested, the RAM isn't sufficient. It just doesn't show this error right away, like you said.

[MFitl]

Originally Posted by Bal21

So, i was able to create a mesh in star ccm in windows*using a .java file. However, when I tried to do the same in linux*system, there was an error saying that the macro file couldn't be opened

*Any thoughts?

Take a look into the filenames in your JAVA-script. There is a difference in the path naming

Linux: /home/username/starccm/testcase.sim
Windows: C:\User\username\starccm\testcase.sim

And yes RAM could be an issue, set the mesh to coareser values.