The potential of GPUs for VLSI physical design automation

Graphic processing units (GPUs) have evolved to provide superb floating-point computing throughput, while parallelization will be the next major milestone for the development of electronic design automation technology. This work aims to harvesting the computing power for the VLSI physical design automation applications. We¿re developing efficient, parallel computing kernels on the CUDA platform for key physical design tools including cell placer, router, and static timing analyzer. The early results show encouraging potentials of GPUs for parallel implementation of EDA algorithms.