A Novel Reduction Circuit Based on Binary Tree Path Partition on FPGAs

Due to high parallelism, field-programmable gate arrays are widely used as accelerators in engineering and scientific fields, which involve a large number of operations of vector and matrix. High-performance accumulation circuits are the key to large-scale matrix operations. By selecting the adder a...

Full description

Bibliographic Details
Main Authors: Linhuai Tang, Zhihong Huang, Gang Cai, Yong Zheng, Jiamin Chen
Format: Article
Language:English
Published: MDPI AG 2021-01-01
Series:Algorithms
Subjects:
Online Access:https://www.mdpi.com/1999-4893/14/2/30
id doaj-3f677b73c9ed43b3a1b5bf50c97eb2ea
record_format Article
spelling doaj-3f677b73c9ed43b3a1b5bf50c97eb2ea2021-01-21T00:06:27ZengMDPI AGAlgorithms1999-48932021-01-0114303010.3390/a14020030A Novel Reduction Circuit Based on Binary Tree Path Partition on FPGAsLinhuai Tang0Zhihong Huang1Gang Cai2Yong Zheng3Jiamin Chen4Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, ChinaAerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, ChinaAerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, ChinaAerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, ChinaAerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, ChinaDue to high parallelism, field-programmable gate arrays are widely used as accelerators in engineering and scientific fields, which involve a large number of operations of vector and matrix. High-performance accumulation circuits are the key to large-scale matrix operations. By selecting the adder as the reduction operator, the reduction circuit can implement the accumulation function. However, the pipelined adder will bring challenges to the design of the reduction circuit. To solve this problem, we propose a novel reduction circuit based on binary tree path partition, which can simultaneously handle multiple data sets with arbitrary lengths. It divides the input data into multiple groups and sends them to different iterations for calculation. The elements belonging to the same data set in each group are added to obtain a partial result, and the partial results of the same data set are added to achieve the final result. Compared with other reduction methods, it has the least area-time product.https://www.mdpi.com/1999-4893/14/2/30pipelinevector reductionaccumulatorFPGAs
collection DOAJ
language English
format Article
sources DOAJ
author Linhuai Tang
Zhihong Huang
Gang Cai
Yong Zheng
Jiamin Chen
spellingShingle Linhuai Tang
Zhihong Huang
Gang Cai
Yong Zheng
Jiamin Chen
A Novel Reduction Circuit Based on Binary Tree Path Partition on FPGAs
Algorithms
pipeline
vector reduction
accumulator
FPGAs
author_facet Linhuai Tang
Zhihong Huang
Gang Cai
Yong Zheng
Jiamin Chen
author_sort Linhuai Tang
title A Novel Reduction Circuit Based on Binary Tree Path Partition on FPGAs
title_short A Novel Reduction Circuit Based on Binary Tree Path Partition on FPGAs
title_full A Novel Reduction Circuit Based on Binary Tree Path Partition on FPGAs
title_fullStr A Novel Reduction Circuit Based on Binary Tree Path Partition on FPGAs
title_full_unstemmed A Novel Reduction Circuit Based on Binary Tree Path Partition on FPGAs
title_sort novel reduction circuit based on binary tree path partition on fpgas
publisher MDPI AG
series Algorithms
issn 1999-4893
publishDate 2021-01-01
description Due to high parallelism, field-programmable gate arrays are widely used as accelerators in engineering and scientific fields, which involve a large number of operations of vector and matrix. High-performance accumulation circuits are the key to large-scale matrix operations. By selecting the adder as the reduction operator, the reduction circuit can implement the accumulation function. However, the pipelined adder will bring challenges to the design of the reduction circuit. To solve this problem, we propose a novel reduction circuit based on binary tree path partition, which can simultaneously handle multiple data sets with arbitrary lengths. It divides the input data into multiple groups and sends them to different iterations for calculation. The elements belonging to the same data set in each group are added to obtain a partial result, and the partial results of the same data set are added to achieve the final result. Compared with other reduction methods, it has the least area-time product.
topic pipeline
vector reduction
accumulator
FPGAs
url https://www.mdpi.com/1999-4893/14/2/30
work_keys_str_mv AT linhuaitang anovelreductioncircuitbasedonbinarytreepathpartitiononfpgas
AT zhihonghuang anovelreductioncircuitbasedonbinarytreepathpartitiononfpgas
AT gangcai anovelreductioncircuitbasedonbinarytreepathpartitiononfpgas
AT yongzheng anovelreductioncircuitbasedonbinarytreepathpartitiononfpgas
AT jiaminchen anovelreductioncircuitbasedonbinarytreepathpartitiononfpgas
AT linhuaitang novelreductioncircuitbasedonbinarytreepathpartitiononfpgas
AT zhihonghuang novelreductioncircuitbasedonbinarytreepathpartitiononfpgas
AT gangcai novelreductioncircuitbasedonbinarytreepathpartitiononfpgas
AT yongzheng novelreductioncircuitbasedonbinarytreepathpartitiononfpgas
AT jiaminchen novelreductioncircuitbasedonbinarytreepathpartitiononfpgas
_version_ 1724330228839350272