Thomas-Ulrich
Thomas-Ulrich
More info: 1. This is not related with Local time stepping (same bug using GTS). 2. The thread async mode works fine. @uphoffc the setup is here: /gpfs/work/pr45fi/di73yeq3/SulawesiSetupBugAsync
This might be related to: https://seissol.readthedocs.io/en/latest/known-issues.html?highlight=hdf5#asynchronous-output use posix instead of hdf5 when using asynchronous output!
Note that this commit from the end of July (3b76efffe53a5475e884fdd934fc4f7e4bdbeb90) gives me a similar error, so probably not related to something recently committed. Edit: note also that the error is...
Ah yes the wavefield output does not work with single precision. While i have further reason to think that there is a problem either with the setup or seissol I...
I reopen the issue. Now I check nans with a receiver at the source location. I get the nan from t>0.
Hi, **short version:** I found a work-around for running this setup with single precision: recentering the mesh allow recovering stability with the same CFL as double precision. **long version:** I...
This is an almost perfectly documented problem: https://github.com/SeisSol/SeisSol/issues/80
good to know. Maybe we can document this idea and close the issue?
I can't reproduce your issue with the latest master, build with seissol-env. (https://seissol.readthedocs.io/en/latest/supermuc.html)
Hi, I m accusing #281 (without proof). I also always use posix, but I think some of my colleagues used hdf5 without major issues. For me at least, I add...