protty

Would also like to recommend [semaphore based synchronization primitives](https://preshing.com/20150316/semaphores-are-surprisingly-versatile/) if it hasn't been considered already, and if simplicity (but not necessarily throughput) is a top priority. Related, but for a...

Any further opinions on this?

> Thanks @kprotty for setting this up. Thanks @jorangreef for the detailed response! This is an area where I strongly believe zig should make some good decisions after having interacted...

> IO functions would take an event loop @vlada-dudr The goal of the discussion is to shed more light on the "hows" and "whys" of asynchronous computing. This also leads...

@vlada-dudr For the "hows" and "whys" part, I meant it as context to transition into why I don't think passing an "event loop" around could be equated to the Allocator...

@kvelicka > opting for a linux-centric event loop design FWIW `io_uring` isn't necessarily linux-centric, but made popular by linux since it's more flexible. The design of "batching" and "asynchronous" syscalls...

> I don't think you mentioned OS X in your response, not sure what the situation with a io_uring-like approach is there Darwin (macos, ios, etc.) uses [kqueue](https://en.wikipedia.org/wiki/Kqueue), an event...

@locriacyber Message passing is an easy to grok, unified communication scheme. But it's not mechanically efficient given memory or serialization overhead it implies. Many times, shared memory is either more...

> There is no cost if you only pass pointers around. @locriacyber The cost is that the pointers have to be escaped as they no longer have a static/linear lifetime...

It's been a while, but wanted to check in on this again. Would the following be a good consensus? * ~~Remove the concept of `io_mode`~~ keep `io_mode`. * Single-Threaded event...