Optimizing data intensive GPGPU computations for DNA sequence alignment

Cole Trapnell; Michael C. Schatz

doi:10.1016/j.parco.2009.05.002

Optimizing data intensive GPGPU computations for DNA sequence alignment

Cole Trapnell
, Michael C. Schatz

Research output: Contribution to journal › Article › peer-review

Abstract

MUMmerGPU uses highly-parallel commodity graphics processing units (GPU) to accelerate the data-intensive computation of aligning next generation DNA sequence data to a reference sequence for use in diverse applications such as disease genotyping and personal genomics. MUMmerGPU 2.0 features a new stackless depth-first-search print kernel and is 13× faster than the serial CPU version of the alignment code and nearly 4× faster in total computation time than MUMmerGPU 1.0. We exhaustively examined 128 GPU data layout configurations to improve register footprint and running time and conclude higher occupancy has greater impact than reduced latency. MUMmerGPU is available open-source at http://www.mummergpu.sourceforge.net.

Original language	English (US)
Pages (from-to)	429-440
Number of pages	12
Journal	Parallel Computing
Volume	35
Issue number	8-9
DOIs	https://doi.org/10.1016/j.parco.2009.05.002
State	Published - Aug 2009
Externally published	Yes

Keywords

CUDA
GPGPU
Short read mapping
Suffix trees

ASJC Scopus subject areas

Software
Theoretical Computer Science
Hardware and Architecture
Computer Networks and Communications
Computer Graphics and Computer-Aided Design
Artificial Intelligence

Access to Document

10.1016/j.parco.2009.05.002

Cite this

@article{181aeaddd5ce47e4af4d187c46b9e48c,

title = "Optimizing data intensive GPGPU computations for DNA sequence alignment",

abstract = "MUMmerGPU uses highly-parallel commodity graphics processing units (GPU) to accelerate the data-intensive computation of aligning next generation DNA sequence data to a reference sequence for use in diverse applications such as disease genotyping and personal genomics. MUMmerGPU 2.0 features a new stackless depth-first-search print kernel and is 13× faster than the serial CPU version of the alignment code and nearly 4× faster in total computation time than MUMmerGPU 1.0. We exhaustively examined 128 GPU data layout configurations to improve register footprint and running time and conclude higher occupancy has greater impact than reduced latency. MUMmerGPU is available open-source at http://www.mummergpu.sourceforge.net.",

keywords = "CUDA, GPGPU, Short read mapping, Suffix trees",

author = "Cole Trapnell and Schatz, \{Michael C.\}",

note = "Funding Information: The authors would like to thank Steven Salzberg from the University of Maryland for helpful comments on the manuscript. This work was supported in part by National Institutes of Health Grants R01-LM006845 and R01-GM083873.",

year = "2009",

month = aug,

doi = "10.1016/j.parco.2009.05.002",

language = "English (US)",

volume = "35",

pages = "429--440",

journal = "Parallel Computing",

issn = "0167-8191",

publisher = "Elsevier B.V.",

number = "8-9",

}

TY - JOUR

T1 - Optimizing data intensive GPGPU computations for DNA sequence alignment

AU - Trapnell, Cole

AU - Schatz, Michael C.

N1 - Funding Information: The authors would like to thank Steven Salzberg from the University of Maryland for helpful comments on the manuscript. This work was supported in part by National Institutes of Health Grants R01-LM006845 and R01-GM083873.

PY - 2009/8

Y1 - 2009/8

N2 - MUMmerGPU uses highly-parallel commodity graphics processing units (GPU) to accelerate the data-intensive computation of aligning next generation DNA sequence data to a reference sequence for use in diverse applications such as disease genotyping and personal genomics. MUMmerGPU 2.0 features a new stackless depth-first-search print kernel and is 13× faster than the serial CPU version of the alignment code and nearly 4× faster in total computation time than MUMmerGPU 1.0. We exhaustively examined 128 GPU data layout configurations to improve register footprint and running time and conclude higher occupancy has greater impact than reduced latency. MUMmerGPU is available open-source at http://www.mummergpu.sourceforge.net.

AB - MUMmerGPU uses highly-parallel commodity graphics processing units (GPU) to accelerate the data-intensive computation of aligning next generation DNA sequence data to a reference sequence for use in diverse applications such as disease genotyping and personal genomics. MUMmerGPU 2.0 features a new stackless depth-first-search print kernel and is 13× faster than the serial CPU version of the alignment code and nearly 4× faster in total computation time than MUMmerGPU 1.0. We exhaustively examined 128 GPU data layout configurations to improve register footprint and running time and conclude higher occupancy has greater impact than reduced latency. MUMmerGPU is available open-source at http://www.mummergpu.sourceforge.net.

KW - CUDA

KW - GPGPU

KW - Short read mapping

KW - Suffix trees

UR - https://www.scopus.com/pages/publications/69949100622

UR - https://www.scopus.com/pages/publications/69949100622#tab=citedBy

U2 - 10.1016/j.parco.2009.05.002

DO - 10.1016/j.parco.2009.05.002

M3 - Article

AN - SCOPUS:69949100622

SN - 0167-8191

VL - 35

SP - 429

EP - 440

JO - Parallel Computing

JF - Parallel Computing

IS - 8-9

ER -

Optimizing data intensive GPGPU computations for DNA sequence alignment

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this