ARL中Gridding算法的并行化实现
|
摘要
针对海量天文数据实时性处理效率低的问题,通过对SKA图像采集及成像ARL算法库中耗时较长的Gridding算法进行耗时分析,找出了该算法中调用频率高且运行时间长的两个函数convolutional-grid和convolutional-degrid,利用GPU的多线程并行化处理降低两个函数的循环迭代,实现了Gridding算法在GPU和CPU上的协同运行.验证实验结果表明,在相同的数据量下,改进后的Gridding算法运行时间大大缩短,特别是在处理海量数据时,有效提高了ARL的整体运行效率.
Aiming at the low real-time processing efficiency of massive astronomical data,through time-consuming analysis of gridding algorithm in SKA image acquisition and imaging ARL library,two functions of convolutional-grid and convolutional-degrid with high frequency and long running time were found out in this algorithm. Then,two functions were parallelized on GPU by multi-threading to realize the cooperative operation of gridding algorithm on GPU and CPU. The experimental results showed that under the same amount of data,the running time of the improved gridding algorithm was greatly shortened,especially when dealing with massive data,the overall running efficiency of ARL was effectively improved.
Aiming at the low real-time processing efficiency of massive astronomical data,through time-consuming analysis of gridding algorithm in SKA image acquisition and imaging ARL library,two functions of convolutional-grid and convolutional-degrid with high frequency and long running time were found out in this algorithm. Then,two functions were parallelized on GPU by multi-threading to realize the cooperative operation of gridding algorithm on GPU and CPU. The experimental results showed that under the same amount of data,the running time of the improved gridding algorithm was greatly shortened,especially when dealing with massive data,the overall running efficiency of ARL was effectively improved.
引文
[1]RAZAVI-GHODS N,ACEDO E D L,EL-MAKA-DEMA A,et al.Analysis of sky contributions to system temperature for low frequency SKA aperture array geometries[J].Experimental Astronomy,2012,33(1):141.
[2]ZHANG Y,BROWN A K.Bunny ear combline antennas for compact wide-band dual-polarized aperture array[J].IEEE Transactions on Antennas and Propagation,2011,59(8):3071.
[3]彭晓明,郭浩然,庞建民.多核处理器---技术、趋势和挑战[J].计算机科学,2012,39(S3):320.
[4]BROUW W N.Aperture synthesis[J].Methods in Computational Physics,1975,14:131.
[5]O’SULLIVAN J D.A fast sinc function gridding algorithm for fourier inversion in computer tomography[J].IEEE Transactions on Medical Imaging,1985,4(4):200.
[6]MEYER C H,HU B S,NISHIMURA D G,et al.Fast spiral coronary artery imaging[J].Magnetic Resonance in Medicine,1992,28(2):202.
[7]劳保强,王俊义,王锦清,等.基于卷积核网格化二维近程微波全息[J].微波学报,2014,30(5):82.
[8]VARBANESCU A L.On the effective parallel programming of multi-core processors[D].Romania:Universitatea Politehnica Bucuresti,2010.
[9]CHENG J.CUDA by example:an introduction to general-purpose GPU programming[M].Boston:Addison-Wesley Professional,2010.
[2]ZHANG Y,BROWN A K.Bunny ear combline antennas for compact wide-band dual-polarized aperture array[J].IEEE Transactions on Antennas and Propagation,2011,59(8):3071.
[3]彭晓明,郭浩然,庞建民.多核处理器---技术、趋势和挑战[J].计算机科学,2012,39(S3):320.
[4]BROUW W N.Aperture synthesis[J].Methods in Computational Physics,1975,14:131.
[5]O’SULLIVAN J D.A fast sinc function gridding algorithm for fourier inversion in computer tomography[J].IEEE Transactions on Medical Imaging,1985,4(4):200.
[6]MEYER C H,HU B S,NISHIMURA D G,et al.Fast spiral coronary artery imaging[J].Magnetic Resonance in Medicine,1992,28(2):202.
[7]劳保强,王俊义,王锦清,等.基于卷积核网格化二维近程微波全息[J].微波学报,2014,30(5):82.
[8]VARBANESCU A L.On the effective parallel programming of multi-core processors[D].Romania:Universitatea Politehnica Bucuresti,2010.
[9]CHENG J.CUDA by example:an introduction to general-purpose GPU programming[M].Boston:Addison-Wesley Professional,2010.