An Efficient Memory-Address Remapping Technique for High-Throughput QC-LDPC Decoder

Ji Hwan Yoon, Jongsun Park

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

This paper presents an efficient memory-address remapping technique for a high-throughput quasi-cyclic low-density parity check (QC-LDPC) decoder. In general, an LDPC decoder needs a large size of embedded memories for the temporal storage of the check node process (CNP) and variable node process (VNP) outputs. To increase the decoder throughput, overlapping the CNP and VNP operations is necessary; however, the parallel operations are mainly restricted by the embedded memory bandwidth. This work presents an efficient memory management approach in an LDPC decoder, where the memory-address conflicts and redundant memory-read operations are effectively reduced by using a proposed memory-address remapping technique. As a result, parallel variable node unit operations significantly increase, leading to higher throughput. When the proposed approach is applied to the various code rates of IEEE std. 802.16-2009, increases in decoding speed of up to 1.52X per iteration are achieved for overlapped message passing algorithm-based architecture, along with considerable reductions in the number of memory-read accesses. Using a 0.13-(Formula presented.)m CMOS process, a QC-LDPC decoder with multi-code rates is implemented, and the experimental results show that the proposed decoder achieves considerable throughput area ratio increase with energy savings compared to the conventional approaches.

Original languageEnglish
Pages (from-to)3457-3473
Number of pages17
JournalCircuits, Systems, and Signal Processing
Volume33
Issue number11
DOIs
Publication statusPublished - 2014 Jan 1

Fingerprint

Parity
High Throughput
Throughput
Data storage equipment
Vertex of a graph
Message-passing Algorithms
Memory Management
Energy Saving
Message passing
Decoding
Overlapping
Bandwidth
Energy conservation
Iteration
Unit
Necessary
Output
Experimental Results

Keywords

  • LDPC decoder
  • Low-density parity check (LDPC) codes
  • Overlapped message passing (OMP)
  • Quasi-cyclic (QC) codes

ASJC Scopus subject areas

  • Signal Processing
  • Applied Mathematics

Cite this

An Efficient Memory-Address Remapping Technique for High-Throughput QC-LDPC Decoder. / Yoon, Ji Hwan; Park, Jongsun.

In: Circuits, Systems, and Signal Processing, Vol. 33, No. 11, 01.01.2014, p. 3457-3473.

Research output: Contribution to journalArticle

@article{946cf6ad46c147c6a3c377a989dde53e,
title = "An Efficient Memory-Address Remapping Technique for High-Throughput QC-LDPC Decoder",
abstract = "This paper presents an efficient memory-address remapping technique for a high-throughput quasi-cyclic low-density parity check (QC-LDPC) decoder. In general, an LDPC decoder needs a large size of embedded memories for the temporal storage of the check node process (CNP) and variable node process (VNP) outputs. To increase the decoder throughput, overlapping the CNP and VNP operations is necessary; however, the parallel operations are mainly restricted by the embedded memory bandwidth. This work presents an efficient memory management approach in an LDPC decoder, where the memory-address conflicts and redundant memory-read operations are effectively reduced by using a proposed memory-address remapping technique. As a result, parallel variable node unit operations significantly increase, leading to higher throughput. When the proposed approach is applied to the various code rates of IEEE std. 802.16-2009, increases in decoding speed of up to 1.52X per iteration are achieved for overlapped message passing algorithm-based architecture, along with considerable reductions in the number of memory-read accesses. Using a 0.13-(Formula presented.)m CMOS process, a QC-LDPC decoder with multi-code rates is implemented, and the experimental results show that the proposed decoder achieves considerable throughput area ratio increase with energy savings compared to the conventional approaches.",
keywords = "LDPC decoder, Low-density parity check (LDPC) codes, Overlapped message passing (OMP), Quasi-cyclic (QC) codes",
author = "Yoon, {Ji Hwan} and Jongsun Park",
year = "2014",
month = "1",
day = "1",
doi = "10.1007/s00034-014-9808-3",
language = "English",
volume = "33",
pages = "3457--3473",
journal = "Circuits, Systems, and Signal Processing",
issn = "0278-081X",
publisher = "Birkhause Boston",
number = "11",

}

TY - JOUR

T1 - An Efficient Memory-Address Remapping Technique for High-Throughput QC-LDPC Decoder

AU - Yoon, Ji Hwan

AU - Park, Jongsun

PY - 2014/1/1

Y1 - 2014/1/1

N2 - This paper presents an efficient memory-address remapping technique for a high-throughput quasi-cyclic low-density parity check (QC-LDPC) decoder. In general, an LDPC decoder needs a large size of embedded memories for the temporal storage of the check node process (CNP) and variable node process (VNP) outputs. To increase the decoder throughput, overlapping the CNP and VNP operations is necessary; however, the parallel operations are mainly restricted by the embedded memory bandwidth. This work presents an efficient memory management approach in an LDPC decoder, where the memory-address conflicts and redundant memory-read operations are effectively reduced by using a proposed memory-address remapping technique. As a result, parallel variable node unit operations significantly increase, leading to higher throughput. When the proposed approach is applied to the various code rates of IEEE std. 802.16-2009, increases in decoding speed of up to 1.52X per iteration are achieved for overlapped message passing algorithm-based architecture, along with considerable reductions in the number of memory-read accesses. Using a 0.13-(Formula presented.)m CMOS process, a QC-LDPC decoder with multi-code rates is implemented, and the experimental results show that the proposed decoder achieves considerable throughput area ratio increase with energy savings compared to the conventional approaches.

AB - This paper presents an efficient memory-address remapping technique for a high-throughput quasi-cyclic low-density parity check (QC-LDPC) decoder. In general, an LDPC decoder needs a large size of embedded memories for the temporal storage of the check node process (CNP) and variable node process (VNP) outputs. To increase the decoder throughput, overlapping the CNP and VNP operations is necessary; however, the parallel operations are mainly restricted by the embedded memory bandwidth. This work presents an efficient memory management approach in an LDPC decoder, where the memory-address conflicts and redundant memory-read operations are effectively reduced by using a proposed memory-address remapping technique. As a result, parallel variable node unit operations significantly increase, leading to higher throughput. When the proposed approach is applied to the various code rates of IEEE std. 802.16-2009, increases in decoding speed of up to 1.52X per iteration are achieved for overlapped message passing algorithm-based architecture, along with considerable reductions in the number of memory-read accesses. Using a 0.13-(Formula presented.)m CMOS process, a QC-LDPC decoder with multi-code rates is implemented, and the experimental results show that the proposed decoder achieves considerable throughput area ratio increase with energy savings compared to the conventional approaches.

KW - LDPC decoder

KW - Low-density parity check (LDPC) codes

KW - Overlapped message passing (OMP)

KW - Quasi-cyclic (QC) codes

UR - http://www.scopus.com/inward/record.url?scp=84920972876&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84920972876&partnerID=8YFLogxK

U2 - 10.1007/s00034-014-9808-3

DO - 10.1007/s00034-014-9808-3

M3 - Article

VL - 33

SP - 3457

EP - 3473

JO - Circuits, Systems, and Signal Processing

JF - Circuits, Systems, and Signal Processing

SN - 0278-081X

IS - 11

ER -