Different results between local laptop and the cluster using the exactly same codes

[zeng...@umn.edu]

Hello all,

I am playing with the irregular wave example in IBAMR. I can successfully run it with 4 cores on my laptop. However, when I move the same codes to the cluster (e.g. cori and onyx), the new mass becomes Nan in the first step. And the level set field is also gradually contaminated in the in-let.

The comparisons of the results are listed in the ppt (https://drive.google.com/file/d/1spVG26SDxCV6mxplyZcc3X5vJVZ8NiBF/view?usp=sharing). 

Any suggestions or comments? Not quite sure whether it is a library issue or an example issue. 

Bty, in the cluster (e.g. cori and onyx), I also tested another gravity wave case with the periodic bcs, which worked well. It is only this irregular wave case with the inlet bcs which produces the Nan in the first step.

Best.

[Boyce Griffith]

On Mar 20, 2021, at 2:48 PM, 'zeng...@umn.edu' via IBAMR Developers <ibam...@googlegroups.com> wrote:

Hello all,

I am playing with the irregular wave example in IBAMR. I can successfully run it with 4 cores on my laptop. However, when I move the same codes to the cluster (e.g. cori and onyx), the new mass becomes Nan in the first step. And the level set field is also gradually contaminated in the in-let.

Are you running / can you run with the same number of cores on the cluster as in your laptop? Can you run with debugging enabled to check to see if you happen to be tripping any assertions?

What compilers are you using on the cluster?

The comparisons of the results are listed in the ppt (https://drive.google.com/file/d/1spVG26SDxCV6mxplyZcc3X5vJVZ8NiBF/view?usp=sharing). 

Any suggestions or comments? Not quite sure whether it is a library issue or an example issue. 

Bty, in the cluster (e.g. cori and onyx), I also tested another gravity wave case with the periodic bcs, which worked well. It is only this irregular wave case with the inlet bcs which produces the Nan in the first step.

Best.

--
You received this message because you are subscribed to the Google Groups "IBAMR Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ibamr-dev+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ibamr-dev/9189b218-17d9-4942-a087-dfe4a0a16a60n%40googlegroups.com.

[Amneet Bhalla]

Hmm, strange. We run irregular waves all the time for our projects on XSEDE and SDSU clusters.

Which example and input file is this, and how many cores you used? Kaustubh/Rama can give it a quick run to confirm.

To view this discussion on the web visit https://groups.google.com/d/msgid/ibamr-dev/96936F74-45FE-4DDE-AE3B-0830964CDD35%40gmail.com.

--

--Amneet 

[Ramakrishnan Thirumalaisamy]

Hi,

I have run the examples/wave_tank/ex0 with input2d_CaseA_IrregularWave file on fermi cluster with 4 nodes (32 processors on each node). It seems working fine. I attached the log file here for 5-time steps.  Fluid mass is bounded. Currently, the solver reached 2.2 s on the cluster. 

To view this discussion on the web visit https://groups.google.com/d/msgid/ibamr-dev/CAMETWJ1xxGDRz5j693O%3DU7VMHmAxcKVFnpwV0Ez9e54-FojvnA%40mail.gmail.com.

--

Regards

Ramakrishnan T.

[zeng...@umn.edu]

Hi Ramakrishnan,

Can you tell me the input file you use for the irregular wave case? just want to make sure we use the same parameters.

Also, I assume you did not modify the example.cpp before testing.

Best. 

[Amneet Bhalla]

It’s mentioned in the previous email

“ I have run the examples/wave_tank/ex0 with input2d_CaseA_IrregularWave file”

To view this discussion on the web visit https://groups.google.com/d/msgid/ibamr-dev/9c3f7fa8-bcae-44c3-a08f-bf13fc1719d8n%40googlegroups.com.

--

--Amneet 

[Yadong Zeng]

Thanks. Sorry for missing that. 

Let me re-run this example by myself.

To view this discussion on the web visit https://groups.google.com/d/msgid/ibamr-dev/CAMETWJ2SDrV94xqYCn2AXYycP8HyjRfTZh9MwdWTDypaBwagCA%40mail.gmail.com.

--

Yadong (Jordan) Zeng

Ph.D. Candidate

Department of Mechanical Engineering  

University of Minnesota