Expand description
A ready-made dive_eval::Host for a model-checking run: the run’s random
streams (C5), the level of the state being evaluated, the live counters
and TLCGet("config"), and TLCSet("exit").
The engine owns a RunHost, passes it to the binding stage
(dive_model::BindOptions::host) so that start-up definitions and
ASSUMEs already draw from the run’s start-up stream, and then drives it:
RunHost::expand before expanding a state (TLC reseeds its generator
with fingerprint(state) ^ seed), RunHost::set_level before each
check, RunHost::set_counter for the live counters, and
RunHost::exit_requested after each state.
One RunHost serves one evaluating thread at a time: the stream and the
level are the current worker’s. A parallel engine keeps one per worker.