Premium
PyCLiPSM: Harnessing heterogeneous computing resources on CPUs and GPUs for accelerated digital soil mapping
Author(s) -
Zhang Guiming,
Zhu AXing,
Liu Jing,
Guo Shanxin,
Zhu Yunqiang
Publication year - 2021
Publication title -
transactions in gis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.721
H-Index - 63
eISSN - 1467-9671
pISSN - 1361-1682
DOI - 10.1111/tgis.12730
Subject(s) - geospatial analysis , computer science , cuda , symmetric multiprocessor system , parallelizable manifold , python (programming language) , leverage (statistics) , multi core processor , parallel computing , exploit , graphics , general purpose computing on graphics processing units , rendering (computer graphics) , supercomputer , computer architecture , computational science , distributed computing , operating system , computer graphics (images) , artificial intelligence , cartography , computer security , algorithm , geography
Abstract Digital soil mapping (DSM) at high spatial resolutions over large areas often demands considerable computing power. This study aims to harness the heterogeneous computing resources on multi‐core central processing units (CPUs) and graphics processing units (GPUs) to accelerate DSM by implementing PyCLiPSM, a parallel version of the iPSM (individual predictive soil mapping) algorithm which represents the type of geospatial algorithms that is data‐ and compute‐intensive and highly parallelizable. PyCLiPSM was implemented in Python based on the PyOpenCL parallel programming library, which runs on any operating system and exploits the computing power of both CPUs and GPUs. Experiments show that PyCLiPSM can effectively leverage multi‐core CPUs and GPUs to speed up DSM tasks. PyCLiPSM is open‐source and freely available. Using PyCLiPSM as an example, we advocate implementing parallel geospatial algorithms using the PyOpenCL framework to harness the heterogeneous computing resources available to researchers and practitioners for accelerated geospatial analysis.