search

visit-dav / visit

VisIt - Visualization and Data Analysis for Mesh-based Scientific Data
https://visit.llnl.gov
BSD 3-Clause "New" or "Revised" License
437 stars 116 forks source link

Hydra Mesh MIR produces strange results #19943

Open JustinPrivitera opened 1 week ago

JustinPrivitera commented 1 week ago

I was working with a user to understand discrepancies between two meshes created with HYDRA which are supposed to be the same. We made material volume fraction expressions to explore where the materials were and we noticed behavior that is mystifying.

Below is an image of the plot of the volume fractions for material 4 on the first mesh, which looks as we expect:

image

The second mesh contains the first mesh but extends further in the negative and positive x directions. If we use the clip operator to only see the portion of the second mesh that is supposed to be the same as the first mesh, we see what we would expect, minus the sides of the box, which disappear because of the way clip functions. This is the image on the left.

image

The image on the right is the same mesh and the same variable, but instead of using clip to get the part of the mesh that we want, we use subset selection to get only the 4th and 5th materials, which are only used in this section of the mesh. What we see is completely different even though the spatial extents are the same.

Data is on the CZ in the following tar file: /usr/workspace/visit/bug_data/visit_bug_19943.tar.xz

Run VisIt with the script repro1.py to get the first image, run with repro2.py to get the second image.

This issue was with VisIt 3.4.1 and 3.3.3.

JustinPrivitera commented 1 week ago

Notify Jose Milovich when this issue is resolved.

JustinPrivitera commented 1 week ago

matvf should not become meaningless once MIR has taken place. It should produce a field of 0s and 1s.

JustinPrivitera commented 1 week ago

Or honestly, any zonal variable that is not material dependent ought to look exactly the same before and after MIR.

markcmiller86 commented 1 week ago

Do intensive vs. extensive variables play any role in this?

JustinPrivitera commented 1 week ago

I'm guessing you're asking about material-dependent fields. They do play a role.

There are two things to figure out here:

  1. why is matvf completely broken after MIR/how can we fix it
  2. the HYDRA variable ireg similarly looks super wrong after MIR. Why is this the case?
markcmiller86 commented 1 week ago

Sorry to be terse, I mean this kind of thing.

JustinPrivitera commented 1 week ago

Sorry, so then what exactly do you mean? Are you talking about volume dependent and/or material-dependent variables?