Archived issue #0031849
Modeling Algorithms - Boolean operations produce different results on different systems
Description
We encountered slightly different results of geometry generation processes being performed on systems with different processor architecture (Ivy bridge vs. Skylake, Zen2 etc.).
Attached are our resulting breps produced by DRAW script below (Attachment 4 (BREP) and Attachment 3 (BREP)).
1) Can you confirm these different results?
2) Can you provide some details for this different behavior?
Attached are our resulting breps produced by DRAW script below (Attachment 4 (BREP) and Attachment 3 (BREP)).
1) Can you confirm these different results?
2) Can you provide some details for this different behavior?
Steps to reproduce
pload ALL datadir c:/temp restore Attachment 5 (BREP) restore Attachment 6 (BREP) bsection res s1 s2 save res res.brep
Public activity
6 archived notes
Participants are labeled by their role within this record.
The difference is not greater than in 11th digit in double precision values. I think it can be explained by different numeric processors on different platforms. It is hardly possible to fix this problem.
You are right, but these differences increase with ongoing manipulations of the model.
We got aware of it when a (seemingly valid) shape produced on Ivybridge architecture could not be meshed on any platform using CADExchanger SDK (which is going to be investigated).
For the sake of completeness, the shapes are attached (Attachment 1 (BREP), Attachment 2 (BREP)).
We got aware of it when a (seemingly valid) shape produced on Ivybridge architecture could not be meshed on any platform using CADExchanger SDK (which is going to be investigated).
For the sake of completeness, the shapes are attached (Attachment 1 (BREP), Attachment 2 (BREP)).
Both shapes are meshed with no problems using OCCT BRepMesh algorithm.
Dear Author, could you provide some feedback on this bug?
Dear msv,
cadex confirmed a bug in their visualization mesher causing the failing meshing of the slightly different shape.
Issue could be closed, the information shared here is sufficient (at least for myself).
cadex confirmed a bug in their visualization mesher causing the failing meshing of the slightly different shape.
Issue could be closed, the information shared here is sufficient (at least for myself).
Please close this bug.