Namely, we use a fixed size actor cluster and round robin FSP tasks to it instead of a whole actor per target computation function.
This also speeds up startup since all fsp workers in the cluster are spun up concurrently alongside the normal UI spawning sequence.
Reworks the fsp engine daemon to use new cluster APIs added to
tractorin:Namely, we use a fixed size actor cluster and round robin FSP tasks to it instead of a whole actor per target computation function. This also speeds up startup since all fsp workers in the cluster are spun up concurrently alongside the normal UI spawning sequence.