The Implement Of Independent Component Algorithm On The GPU

Currently, there is a bottleneck of CPU computing power, although there are a lot of multi-core CPU have been invented, but they still unable to meet the requirements of computer processing, so designing a new processing core which can improve the processing power of computer is imperative. While the GPU is came into being in this environment. GPU applies to deal with the data which are high-intensive and parallelized.Now, GPU has been applied to many areas, such as matrix addition and subtraction, matrix multiplication, etc. When the number of data reaches a certain level, the effect of the acceleration of GPU will be very obvious. Today there are many different algorithms to achieve these functions. I have studied the existing research and understand many methods of GPU acceleration, and learnt the basic thinking, in order to achieve the realization of the independent component algorithms on the GPU.There are a large number of independent component algorithm can be processed in parallel, it will waste a lot of time if it is running on the CPU, in order to solve this problem. In this paper, we use the NVIDIA’s CUDA architecture which is based on GPU.using the parallel processing technology of GPU. and achieve the independent component algorithm, thus speeding up the algorithm. In this paper, we apply the technology of GPU parallel computing; run the independent component algorithm successfully on the GPU.According to the operating results, analyze the advantages of high density data on the GPU.