WARNING::WriteVtuFile:Are there really no scalars or vectors

Clearly defined bug reports and their fixes
Post Reply
mark smith
Posts: 194
Joined: 26 Aug 2009, 18:20
Location: Peterborough, England

WARNING::WriteVtuFile:Are there really no scalars or vectors

Post by mark smith » 30 Sep 2016, 18:13

Hi Elmerteam
I'm trying an Elmer build we have compiled with MUMPS & HYPER on Ubuntu 16.04 LTS and I'm getting some problems with the Resultsoutput solver sometimes not writing the output vtu files correctly (the .pvtu file is sometimes incorrect) and wondered if this could be a compilation problem? or is it due to some other issue?

I have a heat+cfd sif input file with most of the solvers disabled except resultoutput & savematerials to speed the time to get the paraview files written, the log file is attached.

If I don't export OMP_NUM_THREADS=16 I also get a warning in the log file but I'm unsure if this is required by ElmerSolver_mpi or not as the vtu files for each partition are written, and when solving the problem with all solvers active the simulation seems to proceed correctly apart from not writing the vtu results properly ;-(
Thanks in advance
Regards
Mark
Attachments
log.txt
log file
(6.98 KiB) Downloaded 70 times

raback
Site Admin
Posts: 3264
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: WARNING::WriteVtuFile:Are there really no scalars or vectors

Post by raback » 01 Oct 2016, 00:09

Hi Mark

I'll study the output later but what I can already say is that the if you have 16 MPI tasks you should not set the number of threads to 16. They refer to Open MP threads. You can basically have a hybrid parallelization but this is something that is only being worked on and I would not expect any performance gains yet. So pure MPI partition is the way to go.

-Peter

Post Reply