Consulting on the KBA parallel Method

[yingchiyu]

Hello! I am interested in the KBA parallel aspect of OPENSN. We know that the philosophy of the KBA algorithm is “simultaneous in angle, successive in quadrants” (Baker and Koch, 1998). but it is almost exclusively applied in orthogonal grids. As shown in the pictures, I set up 16 threads to compute on a 2D C5G7 model(opensn/tutorials/meshing/read_2D_msh.lua).

It can be seen that sometimes the KBA region division is irregular, Instead of living up to the "diagonal scan" I wanted to achieve, it felt like the result was more akin to an artificial partition grid assigned to threads in parallel. I would like to know if the KBA parallel in OPENSN performs the geometrical level diagonal sweep as mentioned in the paper by Baker and Koch, or if it only divides the regions for parallel processing (which might be similar to domain decomposition). Or it was my negligence that led to the wrong test results.
Look forward to your reply and thank you for your help!