Nice! But a bit disappointing that read speed doesn't increase with number of threads. Curious what the bottleneck is here (first thought is it can't be the memory because CHT shows higher performance for reads in the graph). A comparison with a read-only hashmap would be nice here.