Is it possible to do parallel computing using SNL-Delft3D-CEC?

[jhst14]

Hey all!

Me againg, I keep investigating the use of the SNL-Delft3D-CEC, Does the version of this commit can run in parallel? Because I've trying to do it, and I got the error of the mismatch position of the turbine (because when doing the parallel run, the mesh is divided). I am just running an example of a bay using tides only. Thanks for your help! JHST

[jnvn7]

Hi @jhst14, short answer is yes, we can run it in parallel. Could share your case and the log file so I can take a look at the output errors?

[jhst14]

hello yes of course i attach the files Thanks for your help! JHST

---

De: Jessica Nguyen @.> Enviado: viernes, 8 de septiembre de 2023 01:51 p. m. Para: sandialabs/SNL-Delft3D-CEC @.> Cc: jhst14 @.>; Mention @.> Asunto: Re: [sandialabs/SNL-Delft3D-CEC] Is it possible to do parallel computing using SNL-Delft3D-CEC? (Issue #6)

Hi @jhst14https://github.com/jhst14, short answer is yes, we can run it in parallel. Could share your case and the log file so I can take a look at the output errors?

— Reply to this email directly, view it on GitHubhttps://github.com/sandialabs/SNL-Delft3D-CEC/issues/6#issuecomment-1712209104, or unsubscribehttps://github.com/notifications/unsubscribe-auth/BBV25LF4YAHGVHDGMLCEIVLXZOAO5ANCNFSM6AAAAAA4DTPJNE. You are receiving this because you were mentioned.Message ID: @.***>

[jnvn7]

Hi @jhst14, did you forget to attach the files?

[jhst14]

hi no i didnt forget it i attached in a .rar file the .rar file include:

i attached again in this e-mail, the file es named: "delfd turbinas en paralelo.zip" [cid:6dcc7fcb-1f3e-4818-b828-494582023296]

JHST

---

De: Jessica Nguyen @.> Enviado: lunes, 11 de septiembre de 2023 08:38 a. m. Para: sandialabs/SNL-Delft3D-CEC @.> Cc: jhst14 @.>; Mention @.> Asunto: Re: [sandialabs/SNL-Delft3D-CEC] Is it possible to do parallel computing using SNL-Delft3D-CEC? (Issue #6)

Hi @jhst14https://github.com/jhst14, did you forget to attach the files?

— Reply to this email directly, view it on GitHubhttps://github.com/sandialabs/SNL-Delft3D-CEC/issues/6#issuecomment-1714135728, or unsubscribehttps://github.com/notifications/unsubscribe-auth/BBV25LCKFKQL2KMHF6YIS53XZ4WAHANCNFSM6AAAAAA4DTPJNE. You are receiving this because you were mentioned.Message ID: @.***>

[jnvn7]

@jhst14, could you share it via the google drive link? It might have been filtered out by our system.

[jhst14]

yes of course, sorry a was answering by email The files can be downloaded here: https://drive.google.com/drive/folders/16b5LfLQcZhXNkNpeklXVcBNShsNje2zc?hl=es i will give you access Thanks in advance!

[jnvn7]

Thanks @jhst14! I got the files and did some test runs. I made a few changes relating to the relative paths of the input files specified in 'agia_turbine_paralell.mdf'. I asssumed you had the files in one level up (directory) when you run the case?

I also changed 'agia.mdf' to match the *.mdf file name ('agia_turbine_parallel.mdf') in the config_d_hydro.xml and config_flow2d3d.xml.

Other than that, everything seems to be running fine. Though I have not let it run till the end time. Followings are the screenshot that show the time to completion using 1 processor vs 5 processors. How many did you use for your case?

Could you tell me specifically what error message you got? I don't see the log files (usually starting with tri-diag.*) in the shared folder.

[jhst14]

Hi!

Thanks for your time, here is the link (https://drive.google.com/file/d/1Q6_YC2YNkmmr_F3fND3RrfxImX-xAjzj/view?usp=sharing) where the tri-diag files are uploaded. I also uploaded the modified MDF (removing relative paths). What we do to run it is we use the 'correr_paralelo.bat'. The folder we selected (in the bat file) is where we installed Delft3D (our institute bought a version of Delft3D in 2016, there is a file named 'delft3D_version.txt' where you can find the version we are using), it is also where the Delft3D-SNL files were copied (from this commit). We used 8 cores.

Thanks for the help!!!

[jnvn7]

Hi @jhst14,

I see. The issue was previously resolved, but the fix was implemented in a newer version of the code and was never integrated back into this one. I'd suggest you to test the case using the upgraded version (https://github.com/sandialabs/SNL-Delft3D-FM-CEC/tree/v4) instead. In that repo, the turbine module is integrated into both dflowfm (Delft3D FM flexible mesh) and flow2d3d (Delft3D 4 structured).

Let me know if that fixes the issue you have.

[alepaladio]

Hey all!

Thank you for all your answers, I think we are close. I'm also a PhD student (working together with @jhst14 ). We have just compiled this solution (https://github.com/sandialabs/SNL-Delft3D-FM-CEC/tree/v4). But after compiling these we do not find the folder flow2d3d for the structured version. Can you please tell us how we can integrate this on the solution (or how can we run it with the current files), or if we need to download a different/additional version. Thanks in advance!

On the attached image we show you what the compilation generated.

[jnvn7]

@alepaladio Glad that the compilation for the unstructured code (dflowfm) went well.

For this version of Delft3D, the structured program flow2d3d has to be compiled separately as it is not completely transferred to CMake yet. The compilation for this can be done similarly to what you used before for the "older version" i.e. referring to one of the sample build scripts in the 'src' folder.

Let me know if this works.

-JN