"When I use complicated libraries on my quad-core processor the compile and render process takes really long (few hours) while it only uses 25% of my cpu... "
This might in theory be due to exhaustion of your physical RAM, and the computer doing swap thrashing to disk. The IO system cannot feed the data fast enough to the CPU so it sits there waiting around for the disk. This diagnosis can be verified by looking at various tools that show your swap space and memory usage alongside your CPU usage, like sysmon.
"It isn't just a matter of recompiling it. It would need to be re-coded to split rendering tasks between threads. Not easy, and may not even be possible with the libraries that it uses to do the actual rendering. "
This is exactly right, OpenSCAD is built on top of the CGAL geometry system (F6 render), and OpenCSG (F5 render), neither of which are parallellized at all. Some people have been experimenting with parallelizing CGAL, but they haven't gotten very far.
[
www.mendeley.com]
The problem seems to be they dont just have to re-code it, they have to re-invent the algorithms of CGAL... and CGAL has been around for 10+ years, so that's a lot of re-inventing to do.
Theoretically OpenSCAD could use a different back-end library, but that would take a lot of work. First, you have to find a backend library thats parallelized. Is OpenCASCADE parallelized? dont know.
