Efficient nursery sizing for managed languages on multi-core processors with shared caches.