High memory usage during prep of large grids

Open uqngibbo opened this issue 3 years ago • 2 comments

Kyle and I have observed the prep program using >120 GB of RAM when processing his 32M cell BOLT-II bolt grid, and failing due to memory limits on some machines. This seems like it could be improved with some attention to how the heap is used during prep.

Sep 02 '22 05:09 uqngibbo

I've not run into any issues building larger grids when rotationally extruding 2D grids with the makeWedgeGrid tool. I wasn't tracking the RAM usage directly, but I built a 64M cell grid on my workstation without issue. Might help direct heap management efforts.

Sep 02 '22 06:09 Whyborn

Interesting. This was importing SU2 grids, which is probably a different code path.

Sep 02 '22 06:09 uqngibbo