Assert Counterclockwise Orientation of Read Gmsh File #115
Conversation
status: failing
src/test_tri_vert_orientation.cpp
Outdated
| 0; | ||
| ccw_face[face] = (!cw) ? 1 : -1; | ||
| }; | ||
| Omega_h::parallel_for(mesh.nfaces(), find_ccw_face, "find_ccw_face"); |
There was a problem hiding this comment.
If the intention is to simply answer the question "are all faces ccw?" then this can be written as a max reduction and we can avoid storing ccw_face[face].
There was a problem hiding this comment.
If this check is going into the Omega_h constructor I agree.
There was a problem hiding this comment.
PUMI has a very useful 'verify' function that checks for mesh and model classification issues (I'm not sure of the full set of checks offhand). This function seems like a good first step towards an Omega_h version of that.
src/test_tri_vert_orientation.cpp
Outdated
| (faceCoords[2][1] - faceCoords[1][1]) * | ||
| (faceCoords[1][0] - faceCoords[0][0]) > | ||
| 0; | ||
| ccw_face[face] = (!cw) ? 1 : -1; |
There was a problem hiding this comment.
Is there a reason why you use 1 and -1? If I ask the question is_this_counter_clockwise and you tell me -1 it seems less obvious to me than 0 which corresponds to false.
|
I have added the suggested changes here. I am now finding the orientation and making sure all the faces are giving the same orientation for all faces in one parallel reduction. |
|
@Fuad-HH Thank you. Please provide a description of the PR in the text box that currently says " No description provided. ". When this gets merged that will be the merge commit message.
|
It's added now. |
|
@Fuad-HH Given the mesh uploaded with this PR, is the test expected to pass or fail? |
It's failing. |
|
/runtests |
|
Test Result: failure (details) |
There was a problem hiding this comment.
LGTM. Thank you.
Note, the test failures #115 (comment) were due to a problem with our self-hosted runner.
It is to be asserted that the triangles' nodes are given in the counter-clockwise direction in the mesh adjacency information of Omega_h mesh objects. Various computations that use Omega_h depend on this information. But when this
gmshfile is read, it gives clockwise adjacency. It indicates that there is some bug in the gmsh reader.The order has to be always (0,3,4), (3,4,0), or (4,0,3), not the other way around.