4 Conclusions and Future Work
This paper presents a cluster-based multilevel placement
algorithm, which is shown to be effective for
commercial HFPGA devices. Firstly, a Markov cluster
algorithm is used to produce a hierarchical series of the
target circuit net list. Then, a multilevel optimization
framework is used to solve the HFPGA placement
problem. Tests show that the algorithm produces better
results than the HFPGA’s specific placer.
The current optimization focuses on the wire length.
Other critical factors could also be considered, such as
timing and power constraints, to further improve the
FPGA placement efficiency.