A parallel algorithm for timing-driven global routing for standard cells

The timing-driven global routing problem is an extremely important and time consuming phase of any automated layout system. In this paper, by integrating high performance interconnection tree construction, wire-sizing, and switch-able segment channel optimization together, we propose an adaptive timing-driven global routing algorithm which minimizes the timing delay as well as circuit area. Our experiments on MCNC benchmarks show that our timing-driven global routing algorithm reduces the maximum path delays significantly from the global router TimberWolfSC. Based on this adaptive timing-driven global routing algorithm, a parallel algorithm on timing-driven global routing for standard cells is given. This algorithm has been implemented on an 8 processor IBM J-40 shared memory multi-processor by using the Message Passing Interface (MPI). Our experimental results show good speedup and circuit delay results for this parallel algorithm using MCNC benchmark circuits.