Conversation
|
Apologies for the delayed response.
Have you done any benchmarking and seen actual improvement, or is it just your expectation that this improves things in theory? I'm concerned there's a subtle, potentially dangerous bug here. I realize this is an already delayed response, but I'm going to hold off on making changes until I can test this thoroughly. Most of the performance overhead is from excessive binary context switching. For the future 0.8.x I plan to have this entire subsystem rewritten in a compiled language, with much more fine-grained use of locking. I'm expecting this to be unnoticeable on even on low-end hardware at that time. |
This greatly improves booting speed, especially when running a system with many strata