国防科技大学学报
國防科技大學學報
국방과기대학학보
JOURNAL OF NATIONAL UNIVERSITY OF DEFENSE TECHNOLOGY
2013年
6期
138-146
,共9页
生物信息学%RNA二级结构预测%最小自由能%混合加速方法
生物信息學%RNA二級結構預測%最小自由能%混閤加速方法
생물신식학%RNA이급결구예측%최소자유능%혼합가속방법
bioinformatics%RNA secondary structure prediction%minimal free energy model%hybrid acceleration
RNA 二级结构预测是生物信息学领域重要的研究方向，基于最小自由能模型的 Zuker 算法是目前该领域最典型使用最广泛的算法之一。本文基于 CPU ＋GPU 的混合计算平台实现了对 Zuker 算法的并行和加速。根据 CPU 和 GPU 计算性能的差异，通过合理的任务分配策略，实现二者之间的并行协作计算和处理单元间的负载平衡；针对 CPU 和 GPU 的不同硬件特性，对 Zuker 算法在 CPU 和 GPU 上的实现分别采取了不同的并行优化方法，提高了混合加速系统的计算性能。实验结果表明，CPU 处理单元在混合系统中承担了14％以上的计算任务，与传统的多核 CPU 并行方案相比，采用混合并行加速方法可获得15．93的全局加速比；与最优的单纯 GPU 加速方案相比，可获得16％的性能提升，并且该混合计算方案可用于对其它生物信息学序列分析应用的并行和加速。
RNA 二級結構預測是生物信息學領域重要的研究方嚮，基于最小自由能模型的 Zuker 算法是目前該領域最典型使用最廣汎的算法之一。本文基于 CPU ＋GPU 的混閤計算平檯實現瞭對 Zuker 算法的併行和加速。根據 CPU 和 GPU 計算性能的差異，通過閤理的任務分配策略，實現二者之間的併行協作計算和處理單元間的負載平衡；針對 CPU 和 GPU 的不同硬件特性，對 Zuker 算法在 CPU 和 GPU 上的實現分彆採取瞭不同的併行優化方法，提高瞭混閤加速繫統的計算性能。實驗結果錶明，CPU 處理單元在混閤繫統中承擔瞭14％以上的計算任務，與傳統的多覈 CPU 併行方案相比，採用混閤併行加速方法可穫得15．93的全跼加速比；與最優的單純 GPU 加速方案相比，可穫得16％的性能提升，併且該混閤計算方案可用于對其它生物信息學序列分析應用的併行和加速。
RNA 이급결구예측시생물신식학영역중요적연구방향，기우최소자유능모형적 Zuker 산법시목전해영역최전형사용최엄범적산법지일。본문기우 CPU ＋GPU 적혼합계산평태실현료대 Zuker 산법적병행화가속。근거 CPU 화 GPU 계산성능적차이，통과합리적임무분배책략，실현이자지간적병행협작계산화처리단원간적부재평형；침대 CPU 화 GPU 적불동경건특성，대 Zuker 산법재 CPU 화 GPU 상적실현분별채취료불동적병행우화방법，제고료혼합가속계통적계산성능。실험결과표명，CPU 처리단원재혼합계통중승담료14％이상적계산임무，여전통적다핵 CPU 병행방안상비，채용혼합병행가속방법가획득15．93적전국가속비；여최우적단순 GPU 가속방안상비，가획득16％적성능제승，병차해혼합계산방안가용우대기타생물신식학서렬분석응용적병행화가속。
Prediction of ribonucleic acid (RNA)secondary structure remains to be one of the most important research areas in bioinformatics. The Zuker algorithm is one of the most popular methods of free energy minimization for RNA secondary structure prediction.However,general-purpose computers including parallel computers or multi-core computers exhibit parallel efficiency of no more than 50% on Zuker.For this problem,a CPU-GPU hybrid computing system that accelerates the Zuker algorithm applications for RNA secondary structure prediction is proposed.The computing tasks were allocated between CPU and GPU for parallel cooperate execution.Performance differences between the CPU and the GPU in the task-allocation scheme were considered to obtain workload balance.To improve the hybrid system performance,the Zuker algorithm was optimally implemented with special methods for CPU and GPU architecture.A speedup of 15.93 ×over optimized multi-core SIMD CPU implementation and performance advantage of 16% over optimized GPU implementation were shown in the experimental results.More than 14% of the sequences were executed on CPU in the hybrid system.To the best of our knowledge,our implementation combining CPU and GPU is the only accelerator platform implementing the complete Zuker algorithm.Moreover,the hybrid computing system is proven to be promising and applicable to accelerate other bioinformatics applications.
