Modeling of a hardware VLSI placement system: accelerating the simulated annealing algorithm

William Jr. Batts

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: TK7874.75 .B38 2005


An essential step in the automation of electronic design is the placement of the physical components on the target semiconductor die. The placement step presents the opportunity to reduce costs in terms of wire length and performance degradation; however it is compute intensive and is NP-complete in terms of obtaining an optimal solution. As designs have grown in complexity and gate count, obtaining an optimal solution is not feasible due to time to market constraints or sheer compute effort required. Heuristic algorithms allow for efficient but sub-optimal designs to be produced with a reduction in processing time. A widely used algorithm is Simulated Annealing (SA). The goal of this work was to develop a model that would enable an analysis into the feasibility of developing a hardware accelerated placement system which uses SA at its core. The SA heuristic was analyzed for possible improvements in efficiency with focus given to targeting the system for hardware. A solution implementing parallel computing with specialized hardware configurations inside a field programmable gate array (FPGA) was investigated as having the possibility to improve the efficiency of the SA-based algorithm. All supporting subsystems were also described for a hardware accelerated model. A large speedup was analytically shown from both accelerating the critical path of the SA algorithm as well as novel methods of improving SA?s efficiency. As data throughput requirements were not included in this work, the results presented may be optimistic for an overall system speedup. However, the results clearly show that future work is warranted in studying the concept of a hardware accelerated placement system.