Thread-per-core style, with executors pinned to cores/core-regions. IO_URING and DirectIO support, and a bunch of nice utilities to make working with the thread-per-core style and sharding easy, including a neat utility to automatically mesh all your executors together with incoming outgoing channels.