In this paper, we propose a scalable massively-parallel algorithm to solve the general mapping problem in large-scale networks in real-time. The proposed parallel algorithm takes advantage of GPU architecture and launches millions of workers to calculate values on a target network simultaneously. Threads are managed through the SIMT execution model and target values are updated through atomic operations. Our experiments show the proposed algorithm can accomplish network mapping (find importance weights for links in a real-world large-scale shared-mobility network) with more than 2 million weights within 1.82 µs (microsecond-level), which is truly real-time. The algorithm performance suggests that mapping computations may no longer be the bottleneck in highly dynamic network-centered problems, as the computations can be completed faster than the solid state drive (SSD) read access latency. Compared to serial algorithms, the speedup is more than 12,000 times. The proposed algorithm is also scalable. Results on simulated data show that even when the network size grows exponentially, microsecond-level computing performance can still be obtained, and even more than 190,000 times speedup can be achieved. The proposed algorithm can serve as a cornerstone for ultra-fast processing of highly dynamic large-scale networks.