Re: private scope data destruction, or GC; How it works?

[hale812]

Oh, I forgot to tell the OS, it is Win7 x64 SP1 with latest patches.
Octave 4.2.0

The function does not fail. The program executes until more sophisticated
toolbox function is called (to be particular, that is smoothmesh from
OpenEMS). The problem if that it fails only after I use large array math. If
I use element-wise "for i=1:..." kind of math, it does not fail. 
mesh.x = SmoothMeshLines(SmoothMeshLines(mesh.x, conf.max_res),
conf.max_res); It is not the final pass though.

More about the problem:
The OpenEMS is a FDTD EM-simulation system with Octave/Matlab interface.
Its disadvantage is very simple orthogonal mesh. Therefore in order to
simulate large real-life problems, I need a very big nonuniform mesh, that
can not be processed in lifetime.
So i decided to reuse as much mesh nodes as possible. When generating planar
antenna arrays I tessellate the contours with vertices snapping to nearest
existing nodes, that's what you've seen above. It should be done very
accurately and quite quick, because mesh is really huge. I got it working,
but the iterative node-search strategy is not optimal, therefore I try to
implement some equality-based analysis to all the vertices... and the math
used results in this problem.



--
View this message in context: 
http://octave.1599824.n4.nabble.com/private-scope-data-destruction-or-GC-How-it-works-tp4681518p4681522.html
Sent from the Octave - General mailing list archive at Nabble.com.