Hybrid Simulation of Optical Properties for Thin Film Using CUDA on NVIDIA GPUs

[[abstract]]本研究使用以CUDA平行運算自行撰寫之光學摹擬與輔助設計多功能程式，其功能包含1.太陽能選擇性吸收膜之膜厚最佳化、2.反射率最佳膜厚擬合、3.雙面鍍膜之反射率摹擬、4. 超晶格(superlattice)最佳膜厚擬合、5. 多層薄膜反射率以及吸收率之計算。再衡量GPU(Graphics Processors Unit)運算能否提高效能與計算準確度是否值得信任。 我所使用的NVIDIA G100系列GPU相較與Intel i7系列CPU效能提高最多達40倍以上，且計算結果的誤差值小於5%，GPU確實有機會取代CPU做大量運算。與Matlab平台開發的摹擬程式相比更高達200倍效能提升以上。[[abstract]]This study is optical simulation and multifunction program design by CUDA, it contains: 1.Optimization in the thickness of solar selective absorbing film, 2.Reflectivity of optimal thickness fitting, 3.The reflectivity simulation of double-sided coating, 4.The optimal film thickness fitting on superlattice, 5.The reflectivity of multilayer films and the calculation of the absorption rate. Then measure whether the performance of GPU can increase the efficiency and the confidence in the accuracy. Compared with Intel i7 series CPU, the efficiency of NVIDIA G100 series GPU increases more than 40 times, and the effect for relative difference is less than that of 5%. That is to say, GPU can be a replacement of CPU to conduct huge calculation. Compared with the simulation programming of development platform with Matlab 2008 , the efficiency increases up to 200 times