Search
Investigation of Parallel Data Processing Using Hybrid High Performance CPU + GPU Systems and CUDA Streams
Abstract
The paper investigates parallel data processing in a hybrid CPU+GPU(s) system using multiple CUDA streams for overlapping communication and computations. This is crucial for efficient processing of data, in particular incoming data stream processing that would naturally be forwarded using multiple CUDA streams to GPUs. Performance is evaluated for various compute time to host-device communication time ratios, numbers of CUDA streams, for various numbers of threads managing computations on GPUs. Tests also reveal benefits of using CUDA MPS for overlapping communication and computations when using multiple processes. Furthermore, using standard memory allocation on a GPU and Unified Memory versions are compared, the latter including programmer added prefetching. Performance of a hybrid CPU+GPU version as well as scaling across multiple GPUs are demonstrated showing good speed-ups of the approach. Finally, the performance per power consumption of selected configurations are presented for various numbers of streams and various relative performances of GPUs and CPUs.
Citations
-
8
CrossRef
-
0
Web of Science
-
1 0
Scopus
Cite as
Full text
- Publication version
- Accepted or Published Version
- License
- Copyright (2020 Institute of Informatics Slovak Academy of Sciences)
Keywords
Details
- Category:
- Articles
- Type:
- artykuły w czasopismach
- Published in:
-
COMPUTING AND INFORMATICS
no. 39,
pages 510 - 536,
ISSN: 1335-9150 - Language:
- English
- Publication year:
- 2020
- Bibliographic description:
- Czarnul P.: Investigation of Parallel Data Processing Using Hybrid High Performance CPU + GPU Systems and CUDA Streams// COMPUTING AND INFORMATICS -Vol. 39,iss. 3 (2020), s.510-536
- DOI:
- Digital Object Identifier (open in new tab) 10.31577/cai_2020_3_510
- Verified by:
- Gdańsk University of Technology
seen 151 times