Search

To solve the problem that the statistics of the common random number sequence is poor and the real-time of generating random number sequence is worse,combining advantages and disadvantages of each two types of traditional random number algorithm and newly random number algorithm,this paper proposed a new method that was based on the Sobol’s sequence random number algorithm. The new method could be used to increase the randomness and uniformity of random number sequence and could be obtained random number sequence samples having a good statistical. Next the new random number algorithm was implemented on the CUDA platform to improve its parallel acceleration rate. Finally by the statistical test and charts test this paper verified the feasibility of the above methods. Compared to the traditional CPU serial operation,the speedup ratio has reached more than 200 times,when generating 217 random numbers. [ABSTRACT FROM AUTHOR]

Copyright of Computer Measurement & Control is the property of Magazine Agency of Computer Measurement & Control and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Multiple spectrograms WVD needs to estimate noise variance for calculating auto-term support region, ad the weak signal will be overwhelmed by the strong signal. To solve this problem, the proposed method determined the threshold of support region by the dual tone dynamic range of the signal, and revised the amplitude of support region, which not only avoided noise variance estimation but also improved the signal's time-frequency concentration. Simulation results show that the proposed method reduces the interring cross-terms and preserves the details of weak signal. Finally the algorithm is achieved based on compute unified device architecture (CUDA) and the speedup can reach 50 times. [ABSTRACT FROM AUTHOR]

The next generation of survey telescopes would yield measurements of billions of galaxies, which would cause inefficient, high cost when using CPU. To address this problem, this paper proposed using the graphics processing units (GPUs) in processing the universe computing problems. Firstly, it studied two cosmological calculations, the two-point angular correlation function and the aperture mass statistic. Then, it implemented the two algorithms on the GPU by constructing code, and using CUDA. Finally, it compared the calculation speeds with comparable code run on the CPU. Experimental results indicate that the calculation speeds, using GPUs, has been significantly improved comparing with using CPUs. [ABSTRACT FROM AUTHOR]

In order to improve the speed and quality of reconstruction image, making use of multi-core parallel of GPU based on CUDA, this paper actualized three steps of ray-casting which was geometric transform, scene traversal, rendering through programmable graphics hardware, it could reduce the simulation time required. By 3D texture, synchronous traversing mechanism of ray thread element and opacity early termination, it could reduce the time required for the final simulation results without affecting the imaging quality. The experimental results show that, the algorithm can complete reconstruction of digital radiography image of high speed and high quality. [ABSTRACT FROM AUTHOR]

Using spectral analysis to identify biological characteristics, leading to a large amount of data, and the actual applications need to real-time processing. Although partial least squares is the most widely used authentication algorithm, the algorithm can not achieve rea-l-time performance for large-scale data flow. In order to solve this application contradiction, this paper proposed a parallel computing strategy based on NVIDIA CUDA architecture which took graphics processor unit (GPU) with the characteristic of massively parallel computing as a computing device and combined the advantage of GPU memory to achieve the partial least squares algorithm. The results of experiment show that partial least squares algorithm implemented by CUDA used on GPU is 47 times faster than that used on the CPU, thus the algorithm's property will be significantly improved, so the PLS algorithm can be applied in the large-scale data stream processing possible. [ABSTRACT FROM AUTHOR]