Make parallelization work on Windows (minor changes to use the multiprocess package)
Parallelize work for RPs across available cores
Remove R style code/comments 😜
Results have been checked for consistency, but not for correctness (see attached files; the file starting with seq_ is the original results with the sequential loop).
Changes have reduced memory use from a peak of ~8-9 GB to ~5-6GB per core.
While this is an improvement, obviously it is not enough.
To support larger datasets, we will have to make further changes to support windowed reading/processing, and perhaps explore use of dask which XArray has support for.
@artessen has done the bulk of the work separating out the code in a way to make this a command line tool.
I suggest we build on his work and use typer for this, which is simple to use.
For the notebooks, we might want to explore mercury
This PR includes changes to:
Results have been checked for consistency, but not for correctness (see attached files; the file starting with
seq_
is the original results with the sequential loop).seq_PNG_FL_ADM3_pop_EAI.csv PNG_FL_ADM3_pop_EAI.csv
Changes have reduced memory use from a peak of ~8-9 GB to ~5-6GB per core.
While this is an improvement, obviously it is not enough.
To support larger datasets, we will have to make further changes to support windowed reading/processing, and perhaps explore use of dask which XArray has support for.
@artessen has done the bulk of the work separating out the code in a way to make this a command line tool. I suggest we build on his work and use typer for this, which is simple to use.
For the notebooks, we might want to explore mercury