2010Unpublished venueRequires access

Proceedings of the Fourth Conference on Partitioned Global Address Space Programming Model

Author information unavailable

Open publisher page 10 citations

Abstract

The partitioned global address space programming (PGAS) model has established itself as both a high-productivity and a high-performance approach to parallel computing. It offers ease-of-use, thanks to its global shared address space abstraction, and high performance, thanks to its locality awareness. The use of PGAS model helps in reducing program development time while achieving efficient parallel execution. The name "partitioned global address space" was first introduced by Katherine Yelick and Tarek El- Ghazawi during IPDPS 2002. But the concepts of PGAS have predated its name. The origins of the PGAS model can be traced back to the Unified Parallel C (UPC), Titanium (a Java dialect), and Co-Array Fortran (CAF) parallel programming languages. The DARPA HPCS program has provided additional thrust to the development of new PGAS languages, such as X10 and Chapel, with the specific goal of increasing the productivity of high-performance computing environments. The Partitioned Global Address Space Programming (PGAS) conference is a forum to present and discuss ideas and research developments in the area of the PGAS model, its languages, and applications. Started in 2005, the first conference was held at AHPRC in Minneapolis, MN. The second one was held at The George Washington University in Washington, DC. The third conference was held at The George Washington University's science and technology campus in Ashburn, VA, October 5th-8th, 2009. Finally, this fourth PGAS conference was held at the Morningside Heights Campus of Columbia University in New York, NY, October 12th-15th, 2010. As in the previous conferences, it offered the participants an exciting program with tutorials, panels, keynotes, invited talks, and both contributed and workshop papers in the relevant PGAS research areas. These proceedings are a collection of the papers presented at the fourth PGAS conference. We received 28 submissions to the conference. Those submissions were reviewed by members of the Program Committee and, from this review process, 13 papers were selected for presentation and publication at the conference. In addition, we selected 7 papers for presentation at the PGAS workshop. We also invited 3 keynote speakers and completed the program with a panel.

About this research paper

What this paper is about

The partitioned global address space programming (PGAS) model has established itself as both a high-productivity and a high-performance approach to parallel computing. It offers ease-of-use, thanks to its global shared address space abstraction, and high performance, thanks to its locality awareness. The use of PGAS model helps in reducing program development time while achieving efficient parallel execution. The name "partitioned global address space" was first introduced by Katherine Yelick and Tarek El- Ghazawi during IPDPS 2002. But the concepts of PGAS have predated its name. The origins of the PGAS model can be traced back to the Unified Parallel C (UPC), Titanium (a Java dialect), and Co-Array Fortran (CAF) parallel programming languages. The DARPA HPCS program has provided additional thrust to the development of new PGAS languages, such as X10 and Chapel, with the specific goal of increasing the productivity of high-performance computing environments. The Partitioned Global Address Space Programming (PGAS) conference is a forum to present and discuss ideas and research developments in the area of the PGAS model, its languages, and applications. Started in 2005, the first conference was held at AHPRC in Minneapolis, MN. The second one was held at The George Washington University in Washington, DC. The third conference was held at The George Washington University's science and technology campus in Ashburn, VA, October 5th-8th, 2009. Finally, this fourth PGAS conference was held at the Morningside Heights Campus of Columbia University in New York, NY, October 12th-15th, 2010. As in the previous conferences, it offered the participants an exciting program with tutorials, panels, keynotes, invited talks, and both contributed and workshop papers in the relevant PGAS research areas. These proceedings are a collection of the papers presented at the fourth PGAS conference. We received 28 submissions to the conference. Those submissions were reviewed by members of the Program Committee and, from this review process, 13 papers were selected for presentation and publication at the conference. In addition, we selected 7 papers for presentation at the PGAS workshop. We also invited 3 keynote speakers and completed the program with a panel.

Why it matters

OpenAlex reports 10 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

The partitioned global address space programming (PGAS) model has established itself as both a high-productivity and a high-performance approach to parallel computing. It offers ease-of-use, thanks to its global shared address space abstraction, and high performance, thanks to its locality awareness. The use of PGAS model helps in reducing program development time while achieving efficient parallel execution. The name "partitioned global address space" was first introduced by Katherine Yelick and Tarek El- Ghazawi during IPDPS 2002. But the concepts of PGAS have predated its name. The origins of the PGAS model can be traced back to the Unified Parallel C (UPC), Titanium (a Java dialect), and Co-Array Fortran (CAF) parallel programming languages. The DARPA HPCS program has provided additional thrust to the development of new PGAS languages, such as X10 and Chapel, with the specific goal of increasing the productivity of high-performance computing environments. The Partitioned Global Address Space Programming (PGAS) conference is a forum to present and discuss ideas and research developments in the area of the PGAS model, its languages, and applications. Started in 2005, the first conference was held at AHPRC in Minneapolis, MN. The second one was held at The George Washington University in Washington, DC. The third conference was held at The George Washington University's science and technology campus in Ashburn, VA, October 5th-8th, 2009. Finally, this fourth PGAS conference was held at the Morningside Heights Campus of Columbia University in New York, NY, October 12th-15th, 2010. As in the previous conferences, it offered the participants an exciting program with tutorials, panels, keynotes, invited talks, and both contributed and workshop papers in the relevant PGAS research areas. These proceedings are a collection of the papers presented at the fourth PGAS conference. We received 28 submissions to the conference. Those submissions were reviewed by members of the Program Committee and, from this review process, 13 papers were selected for presentation and publication at the conference. In addition, we selected 7 papers for presentation at the PGAS workshop. We also invited 3 keynote speakers and completed the program with a panel.

Key concepts: Partitioned global address space, Computer science, Fortran, Programming paradigm, Java, Programming language, Parallel computing

Related papers

Back to paper searchBrowse research topicsOriginal source
Proceedings of the Fourth Conference on Partitioned Global Address Space Programming Model — Research Paper | ScholarLens