I can’t help there @Ole_Petter_Hansen. I did a series of runs over the day using the original model from your first post and the data generating script you provided recently:
|N||K||Shards||linreg time (s)||map_rect time (s)||Ratio|
I got a bit fed up with it as the day went on so I skipped some options.
However as you can see only in one case was
map_rect faster. I suspect this is a fluke also since the linear model took much longer than the model immediately above it, I’m going to rerun this linear model to check it.
So on a second run with different random seed the last row linreg changed form 1450ish to 375
- oh the NA is because this was clearly going to take many hours and I didn’t have the patience haha