Parallelization of the channel width search for FPGA routing

As the FPGA becomes more resourceful year by year, the implementation process becomes longer. In particular , the routing process requires a large portion of the implementation, because the routing patterns increase substantially to realize this exibility. Therefore, to reduce the routing process time, we propose a parallel algorithm for the channel-width search in FPGA routing. This algorithm avoids deterioration of cost functions such as delay and routability because each channel-width search is independent. As a result, we achieve a speedup that is 10.56 times the maximum and 4.18 times the average improvement without cost deterioration on 16 processors.