CGAL users discussion list

[vargap <>]

Dear All,

Similarly to the example '51.3.4 Using Variable Sizing Field / Different
Sizing Field for Different Subdomains' provided in the online manual, I'm
trying to mesh a 3D image domain with the c3t3
(Mesh_complex_3_in_triangulation_3) mesher with the cell size is controlled
within several sub-domains defined by a label image. I don't use
optimization.

I have a relative large image domain (~ 800x800x1400 voxel) and in some
sub-domains a very small feature size, I have to prescribe small cell size
(<2) here, while I can use larger cells (~20) elsewhere. Interestingly,
meshing requires reasonable amount of memory (10-20 GB) for certain settings
of the facet and cell sizes, but seems to increase monotonically to
'infinity' (I stop it before the physical limit, which is ~100GB) for a
small change in these parameters (even if it is the increase of the cell
size, which, intuitively, would require less memory). As I don't clearly see
where exactly in the Delaunay refinement process the memory is taken, I
approach the problem empirically and tend to believe that I have to find
'lucky combinations' of these parameters (very probably in relation with the
feature sizes in the label image) for which the meshing is working more
economically.

I'd like to know if it is due to a memory leak, or I do something wrong and
have to set up these parameters 'cleverly' so that I get the case with the
reasonable memory usage.
Have anyone had similar problems or have a suggestion for this case?

All feedback is highly appreciated!
Thank you in advance,
Peter 



--
View this message in context: 
http://cgal-discuss.949826.n4.nabble.com/Memory-consumption-of-Mesh3-tp4656593.html
Sent from the cgal-discuss mailing list archive at Nabble.com.