Accelerating aerial image simulation using improved CPU/GPU collaborative computing

Aerial image simulation is a fundamental problem in advanced lithography for chip fabrication. Since it requires a huge number of mathematical computations, an efficient yet accurate implementation becomes a necessity. In the literature, graphic processing unit (GPU) or multi-core single instruction multiple data (SIMD) CPU has demonstrated its potential for accelerating simulation. However, the combination of GPU and multi-core SIMD CPU was not exploited thoroughly. In this paper, we present and discuss collaborative computing algorithms for the aerial image simulation on multi-core SIMD CPU and GPU. Our improved method achieves up to 160× speedup over the baseline serial approach and outperforms the state-of-the-art GPU-based approach by up to 4× speedup with a hex-core SIMD CPU and Tesla K10 GPU. We show that the performance on the collaborative computing is promising, and the medium-grained task scheduling is suitable for improving the collaborative efficiency.