Research On Acceleration Of Bézier Curve And Surface Generation Algorithm Based On CUDA

Today is a society of data explosion,but also an era of artificial intelligence.In order to deal with a large number of data and to meet the requirements of the processing power of artificial intelligence algorithms,both industry and academia have conducted long-term exploration.Supercomputers,computer clusters,and dedicated neural network chips such as TPU,Cambrian,etc.are the result of the exploration for computing needs.Whether it is supercomputers,computer clusters,dedicated chips,although they can meet people’s requirements for calculate ability,but the shortcomings are equally obvious.Such as system cost and maintenance costs are expensive,energy consumption is also quite amazing,complex programming,and learning curve is steeper.Dedicated chips also have a relatively fixed function,often only to meet people’s needs for a particular type of computing.So the above three solutions are not suitable for personal and small laboratory.GPU-CPU heterogeneous computing can meet certain computing power under the premise,but also to overcome the shortcomings of the system.Coasting on virtue of versatility,programmability,and computing power advantage,NVIDIA released CUDA programming model will undoubtedly become the industry leader in the GPU-CPU heterogeneous computing aspects.Therefore,CUDA-based application research has become a hot topic in academia and industry research.This paper is a study based on CUDA.Work content:(1)This paper makes a brief analysis of the current research on the application of GPU-CPU heterogeneous technology.The development of GPU-CPU heterogeneous computing technology and the historical background are introduced.analyzed and studied the difference between CPU and GPU,and summarized the fields of CPU and GPU.Introduced the NVIDIA CUDA programming architecture in detail,analyzed the characteristics of the CUDA architecture and the technology for parallelization of the CUDA architecture.(2)The theory and generation algorithm of Bezier curve surface are introduced.According to the Characteristics of Curve Surface Generation Algorithm for Large Number of Parallel Computing and the techniques of CUDA to optimize data parallelism,and generate the curve surface point independence,a method of accelerating the curve and surface algorithm based on CUDA heterogeneous system is proposed.(3)According to the proposed method to generate the curve surface,the Bezier curve surface generation algorithm is transplanted to the CUDA heterogeneous computing platform,and the results are analyzed.