imputio is a scratch workspace for implementing a toy async runtime.
It is intended for enabling experimentation with different scheduler/
executor designs, and is a vehicle for learning more about the internals of asynchronous runtimes. I am interested in the performance and complexity tradeoffs of different designs, for example:
- runtimes using kernel-backed datastructures like epoll or io_uring
- runtimes using multithreading versus thread-per-core
- scheduling algorithms like priority queuing, batch processing, workstealing, etc.
Lots of inspiration has been derived from smol, glommio, tokio-uring, and all the other runtimes out there.