|
Assessment of the effect of processes and threads affinity in IBM POWER computing systems on the parallel applications performance
S. I. Malkovskya, A. A. Sorokina, G. I. Tsoya, V. Yu. Chernykha, K. I. Volovichb a Computing Center of the Far Eastern Branch of the Russian Academy of Sciences, 65 Kim U Chen Str., Khabarovsk 680000, Russian Federation
b Institute of Informatics Problems, Federal Research Center "Computer Sciences and Control" of the Russian Academy of Sciences, 44-2 Vavilov Str., Moscow
119133, Russian Federation
Abstract:
The article is devoted to the study of the performance of computing systems based on modern IBM POWER processors when running parallel applications. The effect of different methods of distributing computational processes and threads among central processors' cores on the efficiency of executing programs developed using MPI (Message Passing Interface) and OpenMP (Open Multi-Processing) technologies is studied. The results obtained can be useful for evaluating the efficiency of algorithms for the distribution of computing resources and the organization of the computing process in distributed heterogeneous computing systems.
Keywords:
heterogeneous computing, IBM POWER8, IBM POWER9, simultaneous multithreading, threads affinity, processes affinity, OpenMP, MPI, NAS Parallel Benchmark.
Received: 12.08.2020
Citation:
S. I. Malkovsky, A. A. Sorokin, G. I. Tsoy, V. Yu. Chernykh, K. I. Volovich, “Assessment of the effect of processes and threads affinity in IBM POWER computing systems on the parallel applications performance”, Sistemy i Sredstva Inform., 31:1 (2021), 97–110
Linking options:
https://www.mathnet.ru/eng/ssi752 https://www.mathnet.ru/eng/ssi/v31/i1/p97
|
|