A GPU-based full-chip source-mask optimization solution

Abstract

A simultaneous optimization of source and mask with full-chip capability is presented. To provide full-chip processing capability, the solution is intentionally based on GPUs as well as CPUs and made scalable to large clusters while maintaining convergence. The approach uses a proprietary search algorithm to converge to an optimal solution in the sense of print quality maximization while obeying existing mask manufacturing, lithography equipment and process technology constraints. The solution is based on a proprietary optimization function that is applicable to both binary and phase shift masks.