My process has several long-running components.
Each component uses the Io interface only for synchronization and timeouts.
For this kind of functionality, each component creates its own Io using .init_single_threaded.
I think the “cost” of each single-threaded Io is negligible, at least in terms of memory consumption.
I also hope it results in fewer locks and less contention.
What is your opinion?
Why not use a single Threaded instance.
it won’t spawn threads unless you io.async or io.concurrent, and you can set limits on those when creating the instance.
did you mean - one shared Threaded for all components
or Threaded per-component with specific init options?
1 across all components
Component
Reasons not to use own single-threaded Threaded Io - if such usage is prohibited or has performance implications
I don’t think that it has problem with performance
Use std.Io.Threaded.global_single_threaded.io() then.
looks as happy path for all pre 0.16 sync. primitives
it worth to publish it as pattern/idiom