Using the FusionSC field tracer seems to cause alot of memory consumption in the tracing (not the tracing for J). I know that it is this line
|
coords = field_tracer.follow_field_lines(pol.R, pol.Z, y_slices, rtol=rtol)[ |
|
1:, ... |
|
] |
This adds a constant memory every iteration. This can range from only 25 MB to multiple GB for very large grids. I tried some things to maybe delete all the views that might keep the array allocated, but have not found any fix using this.
My workaround right now is to also use chunking for this process.
Using the FusionSC field tracer seems to cause alot of memory consumption in the tracing (not the tracing for J). I know that it is this line
zoidberg/zoidberg/zoidberg.py
Lines 178 to 180 in 45e3941
This adds a constant memory every iteration. This can range from only 25 MB to multiple GB for very large grids. I tried some things to maybe delete all the views that might keep the array allocated, but have not found any fix using this.
My workaround right now is to also use chunking for this process.