pMapper: Automatic mapping of parallel Matlab programs

Nadya Bliss, Henry Hoffmann, Robert Bond, Hector Chan, Jeremy Kepner, Edmund Wong

Research output: Chapter in Book/Report/Conference proceedingConference contribution

11 Citations (Scopus)

Abstract

Algorithm implementation efficiency is key to delivering high-performance computing capabilities to demanding, high throughput DoD signal and image processing applications and simulations. Significant progress has been made in compiler optimization of serial programs, but many applications require parallel processing, which brings with it the difficult task of determining efficient mappings of algorithms to multiprocessor computers. The pMapper infrastructure addresses the problem of performance optimization of multistage MATLAB® applications on parallel architectures. pMapper is an automatic performance tuning library written as a layer on top of pMatlab. pMatlab is a parallel Matlab toolbox that provides MATLAB users with global array semantics. While pMatlab abstracts the message-passing interface, the responsibility of generating maps for numerical arrays still falls on the user. A processor map for a numerical array is defined as an assignment of blocks of data to processing elements. Choosing the best mapping for a set of numerical arrays in a program is a nontrivial task that requires significant knowledge of programming languages, parallel computing, and processor architecture. pMapper automates the task of map generation, increasing the ease of programming and productivity. In addition to automating the mapping of parallel Matlab programs, pMapper could be used as a mapping tool for embedded systems. This paper addresses the design details of the pMapper infrastructure and presents preliminary results.

Original languageEnglish (US)
Title of host publicationDepartment of Defense High Performance Computing Modernization Program: Proceedings of the HPCMP Users Group Conference 2005
Pages254-261
Number of pages8
Volume2005
DOIs
StatePublished - 2005
Externally publishedYes
EventDepartment of Defense High Performance Computing Modernization Program: HPCMP Users Group Conference 2005 - Nashville, TN, United States
Duration: Jun 27 2005Jun 30 2005

Other

OtherDepartment of Defense High Performance Computing Modernization Program: HPCMP Users Group Conference 2005
CountryUnited States
CityNashville, TN
Period6/27/056/30/05

Fingerprint

MATLAB
Parallel architectures
Message passing
Parallel processing systems
Processing
Embedded systems
Computer programming languages
Program processors
Signal processing
Image processing
Tuning
Productivity
Semantics
Throughput

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Bliss, N., Hoffmann, H., Bond, R., Chan, H., Kepner, J., & Wong, E. (2005). pMapper: Automatic mapping of parallel Matlab programs. In Department of Defense High Performance Computing Modernization Program: Proceedings of the HPCMP Users Group Conference 2005 (Vol. 2005, pp. 254-261). [1592153] https://doi.org/10.1109/DODUGC.2005.53

pMapper : Automatic mapping of parallel Matlab programs. / Bliss, Nadya; Hoffmann, Henry; Bond, Robert; Chan, Hector; Kepner, Jeremy; Wong, Edmund.

Department of Defense High Performance Computing Modernization Program: Proceedings of the HPCMP Users Group Conference 2005. Vol. 2005 2005. p. 254-261 1592153.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Bliss, N, Hoffmann, H, Bond, R, Chan, H, Kepner, J & Wong, E 2005, pMapper: Automatic mapping of parallel Matlab programs. in Department of Defense High Performance Computing Modernization Program: Proceedings of the HPCMP Users Group Conference 2005. vol. 2005, 1592153, pp. 254-261, Department of Defense High Performance Computing Modernization Program: HPCMP Users Group Conference 2005, Nashville, TN, United States, 6/27/05. https://doi.org/10.1109/DODUGC.2005.53
Bliss N, Hoffmann H, Bond R, Chan H, Kepner J, Wong E. pMapper: Automatic mapping of parallel Matlab programs. In Department of Defense High Performance Computing Modernization Program: Proceedings of the HPCMP Users Group Conference 2005. Vol. 2005. 2005. p. 254-261. 1592153 https://doi.org/10.1109/DODUGC.2005.53
Bliss, Nadya ; Hoffmann, Henry ; Bond, Robert ; Chan, Hector ; Kepner, Jeremy ; Wong, Edmund. / pMapper : Automatic mapping of parallel Matlab programs. Department of Defense High Performance Computing Modernization Program: Proceedings of the HPCMP Users Group Conference 2005. Vol. 2005 2005. pp. 254-261
@inproceedings{c6b5492b33ce41bfbdf5a34c639ac474,
title = "pMapper: Automatic mapping of parallel Matlab programs",
abstract = "Algorithm implementation efficiency is key to delivering high-performance computing capabilities to demanding, high throughput DoD signal and image processing applications and simulations. Significant progress has been made in compiler optimization of serial programs, but many applications require parallel processing, which brings with it the difficult task of determining efficient mappings of algorithms to multiprocessor computers. The pMapper infrastructure addresses the problem of performance optimization of multistage MATLAB{\circledR} applications on parallel architectures. pMapper is an automatic performance tuning library written as a layer on top of pMatlab. pMatlab is a parallel Matlab toolbox that provides MATLAB users with global array semantics. While pMatlab abstracts the message-passing interface, the responsibility of generating maps for numerical arrays still falls on the user. A processor map for a numerical array is defined as an assignment of blocks of data to processing elements. Choosing the best mapping for a set of numerical arrays in a program is a nontrivial task that requires significant knowledge of programming languages, parallel computing, and processor architecture. pMapper automates the task of map generation, increasing the ease of programming and productivity. In addition to automating the mapping of parallel Matlab programs, pMapper could be used as a mapping tool for embedded systems. This paper addresses the design details of the pMapper infrastructure and presents preliminary results.",
author = "Nadya Bliss and Henry Hoffmann and Robert Bond and Hector Chan and Jeremy Kepner and Edmund Wong",
year = "2005",
doi = "10.1109/DODUGC.2005.53",
language = "English (US)",
isbn = "0769524966",
volume = "2005",
pages = "254--261",
booktitle = "Department of Defense High Performance Computing Modernization Program: Proceedings of the HPCMP Users Group Conference 2005",

}

TY - GEN

T1 - pMapper

T2 - Automatic mapping of parallel Matlab programs

AU - Bliss, Nadya

AU - Hoffmann, Henry

AU - Bond, Robert

AU - Chan, Hector

AU - Kepner, Jeremy

AU - Wong, Edmund

PY - 2005

Y1 - 2005

N2 - Algorithm implementation efficiency is key to delivering high-performance computing capabilities to demanding, high throughput DoD signal and image processing applications and simulations. Significant progress has been made in compiler optimization of serial programs, but many applications require parallel processing, which brings with it the difficult task of determining efficient mappings of algorithms to multiprocessor computers. The pMapper infrastructure addresses the problem of performance optimization of multistage MATLAB® applications on parallel architectures. pMapper is an automatic performance tuning library written as a layer on top of pMatlab. pMatlab is a parallel Matlab toolbox that provides MATLAB users with global array semantics. While pMatlab abstracts the message-passing interface, the responsibility of generating maps for numerical arrays still falls on the user. A processor map for a numerical array is defined as an assignment of blocks of data to processing elements. Choosing the best mapping for a set of numerical arrays in a program is a nontrivial task that requires significant knowledge of programming languages, parallel computing, and processor architecture. pMapper automates the task of map generation, increasing the ease of programming and productivity. In addition to automating the mapping of parallel Matlab programs, pMapper could be used as a mapping tool for embedded systems. This paper addresses the design details of the pMapper infrastructure and presents preliminary results.

AB - Algorithm implementation efficiency is key to delivering high-performance computing capabilities to demanding, high throughput DoD signal and image processing applications and simulations. Significant progress has been made in compiler optimization of serial programs, but many applications require parallel processing, which brings with it the difficult task of determining efficient mappings of algorithms to multiprocessor computers. The pMapper infrastructure addresses the problem of performance optimization of multistage MATLAB® applications on parallel architectures. pMapper is an automatic performance tuning library written as a layer on top of pMatlab. pMatlab is a parallel Matlab toolbox that provides MATLAB users with global array semantics. While pMatlab abstracts the message-passing interface, the responsibility of generating maps for numerical arrays still falls on the user. A processor map for a numerical array is defined as an assignment of blocks of data to processing elements. Choosing the best mapping for a set of numerical arrays in a program is a nontrivial task that requires significant knowledge of programming languages, parallel computing, and processor architecture. pMapper automates the task of map generation, increasing the ease of programming and productivity. In addition to automating the mapping of parallel Matlab programs, pMapper could be used as a mapping tool for embedded systems. This paper addresses the design details of the pMapper infrastructure and presents preliminary results.

UR - http://www.scopus.com/inward/record.url?scp=33947689163&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33947689163&partnerID=8YFLogxK

U2 - 10.1109/DODUGC.2005.53

DO - 10.1109/DODUGC.2005.53

M3 - Conference contribution

AN - SCOPUS:33947689163

SN - 0769524966

SN - 9780769524962

VL - 2005

SP - 254

EP - 261

BT - Department of Defense High Performance Computing Modernization Program: Proceedings of the HPCMP Users Group Conference 2005

ER -