Closed Andres-MG closed 1 year ago
This is more complex than we expected. The tests are also failing for other reasons not related to the names of the averaging functions. Some of the control files seem to be wrong, and their names do not always match the ones in the workflow.
I am aware of the naming problem of the control file in the workflow, I'll fix that. What about the other things? All fixed then?
Everything seems to be fine with the last fix that you pushed. The control files were fine, just a silly mistake from my side... Now there is an issue with the VTKHDF module of horses2plt, but I will work on that on a different PR.
@Andres-MG, the error we are getting is weird and doesn't seem to have anything to do with what you changed right?
I think so. It seems that this test has not been run for some time, so I assume this issue is not new. The difference between the expected and the actual values is small, maybe it is not an error...
I would bet it works with ifort and not with gfortran. I think that needs MKL, should it be even run with gfortran?
BTW, it runs in flexo with ifort + mkl
I think that this works with both, maybe the problem is that intel+MKL and gfortran use different implementations that return slightly different results.
There are some inconsistencies in the code regarding the split forms for the incompressible Navier-Stokes equations. With this PR, the split forms called
skew-symmetric-1split
andskew-symmetric-2split
are nowskew-symmetric 1
andskew-symmetric 2
everywhere.