本文關(guān)鍵詞： 功能腦網(wǎng)絡(luò) 網(wǎng)絡(luò)屬性 圖像處理器 統(tǒng)一計算設(shè)備架構(gòu) 加速比　出處：《計算機(jī)工程與設(shè)計》2017年06期 　論文類型：期刊論文

【摘要】：為實(shí)現(xiàn)大規(guī)模功能腦網(wǎng)絡(luò)拓?fù)鋵傩缘母咝в嬎?提出基于GPU并行的腦網(wǎng)絡(luò)屬性分析方法。采用統(tǒng)一計算設(shè)備CUDA架構(gòu),屬性分析方法中的計算密集型操作由GPU完成。以功能腦網(wǎng)絡(luò)為對象,在GPU型號為NVIDIA Quadro K4200的工作站上對該并行方法進(jìn)行模擬,將該方法與基于單程序多數(shù)據(jù)SPMD機(jī)制的腦網(wǎng)絡(luò)屬性分析方法進(jìn)行比較。實(shí)驗(yàn)結(jié)果表明,當(dāng)網(wǎng)絡(luò)節(jié)點(diǎn)數(shù)大于1000時,該方法具有更高的計算性能。
[Abstract]:For the efficient computation of large-scale functional brain networks topology, the brain network properties analysis method is proposed based on parallel GPU. Using CUDA Compute Unified Device Architecture, attribute analysis computation intensive methods of operation is completed by GPU. The functional brain network as the object, in the GPU model NVIDIA Quadro K4200 workstation for the parallel method by simulation, the method and the analysis method of the brain network properties of single program multiple data based on SPMD mechanism were compared. The experimental results show that when the network node number is greater than 1000, this method has higher computing performance.

【作者單位】： 太原理工大學(xué)計算機(jī)科學(xué)與技術(shù)學(xué)院;
【基金】：國家自然科學(xué)基金青年基金項(xiàng)目(61503273) 太原理工大學(xué)校基金項(xiàng)目(1205-04020202)
【分類號】：O157.5
，

本文編號：1529110