摘要
相干体技术在地震勘探资料解释方面得到了广泛的应用,由于相干体技术处理的对象是三维地震数据体,所以算法运算时间较长。为了缩短解释周期,本文充分发挥GPU并行计算优势,对C3相干体算法进行并行化分析。从硬盘读取数据到GPU上计算相干值并写入硬盘的整个过程进行分析,剔除了冗余数据的读取,完成了C3相干体算法的并行化设计与实现。最后分别对串行算法与并行算法进行性能测试,结果表明本文设计的并行算法在保证精度的前提下达到了16倍左右的加速比,对加快地震资料解释具有重要意义。
The coherence technology has been widely used in the interpretation of seismic exploration data.The object of coherence technology processing is three-dimensional seismic data,so the algorithm takes a long time to operate.In order to shorten the interpretation cycle,this paper suggests to fully use the advantages of GPU parallel computation and parallelizes the C3 coherence algorithm.The whole process is analyzed,from data reading of hard disk,calculating coherent values to writing data to hard disk.We eliminate the reading of redundant data,and complete the design and realization of C3 coherence parallel algorithm using the high-performance CUDA technique.Finally,the performance of serial and parallel algorithms are tested.The results show that the parallel algorithm proposed in this paper achieves 16 times the speedup ratio on the premise of ensuring accuracy,and it is of great significance in accelerating interpretation of seismic data.
作者
张全
林柏栎
彭博
王世元
ZHANG Quan;LIN Baiyue;PENG Bo;WANG Shiyuan(School of Computer Science,Southwest Petroleum University,Chengdu,Sichuan 610500;School of Information andCommunication Engineering,University of Electronic Science and Technology of China,Chengdu,Sichuan 611731)
出处
《地质与勘探》
CAS
CSCD
北大核心
2020年第1期147-153,共7页
Geology and Exploration
关键词
地震属性体
断层识别
相干体算法
并行计算
CUDA
GPU
seismic attributes
fault recognition
coherence algorithm
parallel computation
Compute Unified Device Architecture(CUDA)
Graphic Processing Unit(GPU)
