futures. For fine-grained threading this lifts performance to another level. We can now use parallelization in inner loops of Guile programs without impossible overhead.
