异构加速绝对平面检测  

Heterogeneous acceleration absolute flatness testing

作　　者：李雨杭 韩森[1,2] 李雪园 徐春凤 龚晨曦 LI Yuhang;HAN Sen;LI Xueyuan;XU Chunfeng;GONG Chenxi(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China;Suzhou H&L Instruments LLC.,Suzhou 215000,China)

机构地区：[1]上海理工大学光电信息与计算机学院,上海200093 [2]苏州慧利仪器有限责任公司,江苏苏州215000

出　　处：《光学仪器》2025年第1期32-40,共9页Optical Instruments

基　　金：国家重点研发计划(2022YFF0607704)。

摘　　要：光学干涉检测领域的不断发展要求检测仪器具备更高的横向分辨率。高分辨率意味着处理时间变长,测试效率变低。为提高测试效率,提出了一种利用CPU/GPU异构计算并行加速的Zernike多项式绝对平面检测方法。该方法使用CPU进行流程控制,利用GPU多核优势将检测平面中的元素离散并行求解,并在Zernike系数求解中使用混合精度,在峰谷值和均方根值求解中使用线程束原语指令进一步优化性能。结果显示,使用RTX3070-Laptop,在512×512、1024×1024、2048×2048和4096×4096分辨率的光学平面检测中,该方法整体处理速度比CPU处理速度分别提高了近47、56、58和70倍。The growing field of optical interferometry requires instruments with higher lateral resolution.High-resolution implies long processing time,which significantly affects testing efficiency.In order to improve the testing efficiency,a Zernike polynomial absolute flatness testing utilizing parallel acceleration of CPU/GPU heterogeneous computing was proposed,which used CPU for process to control and GPU multi-core advantage to discretize the elements in the flatness for parallel solving.Performance was further optimized useing mixed-precision in Zernike coefficient solving and peak-to-valley(PV)and root mean square(RMS)were solved with warplevel primitive instructions.Using 512×512,1024×1024,2048×2048,and 4096×4096 optically flat surfaces on the RTX3070–Laptop hardware,the speed increased by 47,56,58,and 70 times respectively in the overall process compared to the CPU version.

关 键 词：绝对平面检测 异构并行计算 Zernike多项式拟合 

分 类 号：TN206[电子电信—物理电子学]