performance seems to massively degrade on macOS when large number of spawn_blocking calls are issued, even when the max blocking pool threads are reduced to 8. Using a suggestion by @fnichol to break the deserialization up with a tokio::task::yield_now() call instead.
performance seems to massively degrade on macOS when large number of spawn_blocking calls are issued, even when the max blocking pool threads are reduced to 8. Using a suggestion by @fnichol to break the deserialization up with a tokio::task::yield_now() call instead.