information loss when outputting .vtu using MPI

Shuo LIU

unread,

Feb 14, 2023, 4:59:04 AM2/14/23

to basilisk-fr

Dear all,

I am extracting .vtu files from dump files for post-processing using ParaView with this head file:

http://basilisk.fr/sandbox/acastillo/output_fields/output_vtu_foreach.h

I run with MPI and it generates 128 .vtu files which are combined with a .pvtu file.

The problem is that the parallel results seem to lose some information compared to the results generated by a single core. As you can see from the screenshots I attached, one is a free surface generated with 1 core, and the other is generated with 128 cores.

Does anyone have any relevant experience?

128.png

1.png

Vrionis Yiannis

unread,

Feb 14, 2023, 7:01:50 AM2/14/23

to Shuo LIU, basilisk-fr

I have been facing a similar problem with the view.h header.

When using MPI there seems to be part of the domain lost, whereas when without MPI (even with OpenMP)

the output is as expected.

Cheers,

Yiannis

--
You received this message because you are subscribed to the Google Groups "basilisk-fr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to basilisk-fr...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/basilisk-fr/3eab3c35-17d1-4bdb-b21c-0604e3e0e560n%40googlegroups.com.

Message has been deleted

Maxi Sa

unread,

Feb 15, 2023, 3:57:55 PM2/15/23

to basilisk-fr

Hi,

does it help if you use ParaView's Ghost Cell Generator filter?

Reply all

Reply to author

Forward