Heap Memory Limit

[petrasvestartas]

I am asking here, because I do not know where to ask such a specific question, and you probably know C++ more than anyone else.
We are hitting Heap Memory Error limit when binding CGAL with nanobind for python in azure conda feedstock. Locally building is good. I assume that the cloud building computers have lower memory.

This is the project, which is essentially a cmake project:
https://github.com/compas-dev/compas_cgal

This is the error:
https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=1202333&view=logs&j=535b614b-9e28-51ef-eb67-d0aa83d01d61&t=c0e5205c-8699-555e-ce93-e93093f402f2

Libraries used:
CGAL, Boost, Eigen.

[afabri]

Can you isolate slicer.cpp ?

[petrasvestartas]

I tried to remove 2D polygon skeleton and poisson surface reconstruction and builds without memory issues.

It seems nanobind eats too much RAM memory on windows.
I will try to make nanobind bindings as smaller sets of CGAL methods. During compilation it easily gets above 12 GB RAM.
No ideas where this comes from though, it is a such a small project.

[afabri]

Can you just tell us if you found a workaround for the problem.

[petrasvestartas]

The only solution was to split cpp project into multiple targets.

Nanobind, boost and cgal are heavily templated libraries.
We also removed multithreading for building in conda.
It was always hitting more than 16 gb of RAM.