Skeletons and Asynchronous RPC for Embedded Data and Task Parallel Image Processing

Wouter CAARLS; Pieter JONKER; Henk CORPORAAL

doi:10.1093/ietisy/e89-d.7.2036

Skeletons and Asynchronous RPC for Embedded Data and Task Parallel Image Processing

Wouter CAARLS, Pieter JONKER, Henk CORPORAAL

Full Text Views

0

Cite this

Summary :

Developing embedded parallel image processing applications is usually a very hardware-dependent process, often using the single instruction multiple data (SIMD) paradigm, and requiring deep knowledge of the processors used. Furthermore, the application is tailored to a specific hardware platform, and if the chosen hardware does not meet the requirements, it must be rewritten for a new platform. We have proposed the use of design space exploration [9] to find the most suitable hardware platform for a certain application. This requires a hardware-independent program, and we use algorithmic skeletons [5] to achieve this, while exploiting the data parallelism inherent to low-level image processing. However, since different operations run best on different kinds of processors, we need to exploit task parallelism as well. This paper describes how we exploit task parallelism using an asynchronous remote procedure call (RPC) system, optimized for low-memory and sparsely connected systems such as smart cameras. It uses a futures [16]-like model to present a normal imperative C-interface to the user in which the skeleton calls are implicitly parallelized and pipelined. Simulation provides the task dependency graph and performance numbers for the mapping, which can be done at run time to facilitate data dependent branching. The result is an easy to program, platform independent framework which shields the user from the parallel implementation and mapping of his application, while efficiently utilizing on-chip memory and interconnect bandwidth.

Publication: IEICE TRANSACTIONS on Information Vol.E89-D No.7 pp.2036-2043

Publication Date: 2006/07/01

Publicized

Online ISSN: 1745-1361

DOI: 10.1093/ietisy/e89-d.7.2036

Type of Manuscript: Special Section PAPER (Special Section on Machine Vision Applications)

Category: Parallel and Distributed Computing

Cite this

Copy

Wouter CAARLS, Pieter JONKER, Henk CORPORAAL, "Skeletons and Asynchronous RPC for Embedded Data and Task Parallel Image Processing" in IEICE TRANSACTIONS on Information, vol. E89-D, no. 7, pp. 2036-2043, July 2006, doi: 10.1093/ietisy/e89-d.7.2036.
Abstract: Developing embedded parallel image processing applications is usually a very hardware-dependent process, often using the single instruction multiple data (SIMD) paradigm, and requiring deep knowledge of the processors used. Furthermore, the application is tailored to a specific hardware platform, and if the chosen hardware does not meet the requirements, it must be rewritten for a new platform. We have proposed the use of design space exploration [9] to find the most suitable hardware platform for a certain application. This requires a hardware-independent program, and we use algorithmic skeletons [5] to achieve this, while exploiting the data parallelism inherent to low-level image processing. However, since different operations run best on different kinds of processors, we need to exploit task parallelism as well. This paper describes how we exploit task parallelism using an asynchronous remote procedure call (RPC) system, optimized for low-memory and sparsely connected systems such as smart cameras. It uses a futures [16]-like model to present a normal imperative C-interface to the user in which the skeleton calls are implicitly parallelized and pipelined. Simulation provides the task dependency graph and performance numbers for the mapping, which can be done at run time to facilitate data dependent branching. The result is an easy to program, platform independent framework which shields the user from the parallel implementation and mapping of his application, while efficiently utilizing on-chip memory and interconnect bandwidth.
URL: https://global.ieice.org/en_transactions/information/10.1093/ietisy/e89-d.7.2036/_p

Copy

@ARTICLE{e89-d_7_2036,
author={Wouter CAARLS, Pieter JONKER, Henk CORPORAAL, },
journal={IEICE TRANSACTIONS on Information},
title={Skeletons and Asynchronous RPC for Embedded Data and Task Parallel Image Processing},
year={2006},
volume={E89-D},
number={7},
pages={2036-2043},
abstract={Developing embedded parallel image processing applications is usually a very hardware-dependent process, often using the single instruction multiple data (SIMD) paradigm, and requiring deep knowledge of the processors used. Furthermore, the application is tailored to a specific hardware platform, and if the chosen hardware does not meet the requirements, it must be rewritten for a new platform. We have proposed the use of design space exploration [9] to find the most suitable hardware platform for a certain application. This requires a hardware-independent program, and we use algorithmic skeletons [5] to achieve this, while exploiting the data parallelism inherent to low-level image processing. However, since different operations run best on different kinds of processors, we need to exploit task parallelism as well. This paper describes how we exploit task parallelism using an asynchronous remote procedure call (RPC) system, optimized for low-memory and sparsely connected systems such as smart cameras. It uses a futures [16]-like model to present a normal imperative C-interface to the user in which the skeleton calls are implicitly parallelized and pipelined. Simulation provides the task dependency graph and performance numbers for the mapping, which can be done at run time to facilitate data dependent branching. The result is an easy to program, platform independent framework which shields the user from the parallel implementation and mapping of his application, while efficiently utilizing on-chip memory and interconnect bandwidth.},
keywords={},
doi={10.1093/ietisy/e89-d.7.2036},
ISSN={1745-1361},
month={July},}

Copy

TY - JOUR
TI - Skeletons and Asynchronous RPC for Embedded Data and Task Parallel Image Processing
T2 - IEICE TRANSACTIONS on Information
SP - 2036
EP - 2043
AU - Wouter CAARLS
AU - Pieter JONKER
AU - Henk CORPORAAL
PY - 2006
DO - 10.1093/ietisy/e89-d.7.2036
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E89-D
IS - 7
JA - IEICE TRANSACTIONS on Information
Y1 - July 2006
AB - Developing embedded parallel image processing applications is usually a very hardware-dependent process, often using the single instruction multiple data (SIMD) paradigm, and requiring deep knowledge of the processors used. Furthermore, the application is tailored to a specific hardware platform, and if the chosen hardware does not meet the requirements, it must be rewritten for a new platform. We have proposed the use of design space exploration [9] to find the most suitable hardware platform for a certain application. This requires a hardware-independent program, and we use algorithmic skeletons [5] to achieve this, while exploiting the data parallelism inherent to low-level image processing. However, since different operations run best on different kinds of processors, we need to exploit task parallelism as well. This paper describes how we exploit task parallelism using an asynchronous remote procedure call (RPC) system, optimized for low-memory and sparsely connected systems such as smart cameras. It uses a futures [16]-like model to present a normal imperative C-interface to the user in which the skeleton calls are implicitly parallelized and pipelined. Simulation provides the task dependency graph and performance numbers for the mapping, which can be done at run time to facilitate data dependent branching. The result is an easy to program, platform independent framework which shields the user from the parallel implementation and mapping of his application, while efficiently utilizing on-chip memory and interconnect bandwidth.
ER -