Basic Principles of VLBI FX Type Correlation Algorythms Based on Graphical Processing Units
Transactions of IAA RAS, issue 33, 58–63 (2015)
Keywords: correlator, data processing, high-perfomance computing cluster, HPC, graphical processing unit, GPU
About the paperAbstract
The distinctive feature of the new software correlator which has been designed in IAA RAS is the usage of graphical processing units (GPUs) for fringe stopping, FFT, spectra multiplication, and pcal extracting operations. The near-realtime GPU-based algorithms and software was developed to perform the VGOS data processing. In order to compute 96 GB/s data stream from 6 stations and 2 polarizations algorithms require 56 GPUs and 6 GPUs for pcal extracting.
Citation
V. O. Ken, I. F. Surkis. Basic Principles of VLBI FX Type Correlation Algorythms Based on Graphical Processing Units // Transactions of IAA RAS. — 2015. — Issue 33. — P. 58–63.
@article{ken2015,
abstract = {The distinctive feature of the new software correlator which has been designed in IAA RAS is the usage of graphical processing units (GPUs) for fringe stopping, FFT, spectra multiplication, and pcal extracting operations. The near-realtime GPU-based algorithms and software was developed to perform the VGOS data processing. In order to compute 96 GB/s data stream from 6 stations and 2 polarizations algorithms require 56 GPUs and 6 GPUs for pcal extracting.},
author = {V.~O. Ken and I.~F. Surkis},
issue = {33},
journal = {Transactions of IAA RAS},
keyword = {correlator, data processing, high-perfomance computing cluster, HPC, graphical processing unit, GPU},
pages = {58--63},
title = {Basic Principles of VLBI FX Type Correlation Algorythms Based on Graphical Processing Units},
url = {http://iaaras.ru/en/library/paper/1049/},
year = {2015}
}
TY - JOUR
TI - Basic Principles of VLBI FX Type Correlation Algorythms Based on Graphical Processing Units
AU - Ken, V. O.
AU - Surkis, I. F.
PY - 2015
T2 - Transactions of IAA RAS
IS - 33
SP - 58
AB - The distinctive feature of the new software correlator which has been
designed in IAA RAS is the usage of graphical processing units (GPUs)
for fringe stopping, FFT, spectra multiplication, and pcal extracting
operations. The near-realtime GPU-based algorithms and software was
developed to perform the VGOS data processing. In order to compute 96
GB/s data stream from 6 stations and 2 polarizations algorithms
require 56 GPUs and 6 GPUs for pcal extracting.
UR - http://iaaras.ru/en/library/paper/1049/
ER -