Optimizing Communication by Compression for Multi-GPU Scalable Breadth-First Searches