skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: PaRSEC: A Software Framework for Performance and Productivity on Hybrid, Manycore Platforms

Abstract

As the era of computer architectures dominated by serial processors ends, the convergence of several unprecedented challenges suggests that closing the longstanding "application–architecture performance gap" will become more challenging than ever. To address this problem, the Parallel Runtime Scheduling and Execution Control (PaRSEC) project created a modular software framework that achieved two major objectives: first, it built a task-based runtime capable of delivering portable performance to a wide range of science and engineering applications at all levels of the platform pyramid, including the upcoming 100 Pflop/s systems and then exascale; and second, it supported and facilitated the work of developers in migrating their legacy codes and writing entirely new ones for the emerging hybrid and massively parallel manycore processor system designs. PaRSEC will support multiple domain-specific languages capable of increasing the developers' productivity while also providing the runtime with the constructs and flexibility necessary to exploit the maximal parallelism from parallel applications. Extensive preliminary research in dense linear algebra showed convincingly that a parameterized task graph representation that symbolically describes the algorithm content can achieve the project's twofold objective within that domain. The research also strongly suggested that this powerful method could be generalized to a far-wider variety of applications.

Authors:
 [1]
  1. Univ. of Tennessee, Knoxville, TN (United States)
Publication Date:
Research Org.:
Univ. of Tennessee, Knoxville, TN (United States)
Sponsoring Org.:
USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR) (SC-21)
OSTI Identifier:
1348914
Report Number(s):
DOE-UTK-10682
DOE Contract Number:
SC0010682
Resource Type:
Technical Report
Country of Publication:
United States
Language:
English
Subject:
97 MATHEMATICS AND COMPUTING; task-based runtime; parallel runtime scheduling and execution control; application-architecture performance gap

Citation Formats

Dongarra, Jack. PaRSEC: A Software Framework for Performance and Productivity on Hybrid, Manycore Platforms. United States: N. p., 2016. Web. doi:10.2172/1348914.
Dongarra, Jack. PaRSEC: A Software Framework for Performance and Productivity on Hybrid, Manycore Platforms. United States. doi:10.2172/1348914.
Dongarra, Jack. Thu . "PaRSEC: A Software Framework for Performance and Productivity on Hybrid, Manycore Platforms". United States. doi:10.2172/1348914. https://www.osti.gov/servlets/purl/1348914.
@article{osti_1348914,
title = {PaRSEC: A Software Framework for Performance and Productivity on Hybrid, Manycore Platforms},
author = {Dongarra, Jack},
abstractNote = {As the era of computer architectures dominated by serial processors ends, the convergence of several unprecedented challenges suggests that closing the longstanding "application–architecture performance gap" will become more challenging than ever. To address this problem, the Parallel Runtime Scheduling and Execution Control (PaRSEC) project created a modular software framework that achieved two major objectives: first, it built a task-based runtime capable of delivering portable performance to a wide range of science and engineering applications at all levels of the platform pyramid, including the upcoming 100 Pflop/s systems and then exascale; and second, it supported and facilitated the work of developers in migrating their legacy codes and writing entirely new ones for the emerging hybrid and massively parallel manycore processor system designs. PaRSEC will support multiple domain-specific languages capable of increasing the developers' productivity while also providing the runtime with the constructs and flexibility necessary to exploit the maximal parallelism from parallel applications. Extensive preliminary research in dense linear algebra showed convincingly that a parameterized task graph representation that symbolically describes the algorithm content can achieve the project's twofold objective within that domain. The research also strongly suggested that this powerful method could be generalized to a far-wider variety of applications.},
doi = {10.2172/1348914},
journal = {},
number = ,
volume = ,
place = {United States},
year = {Thu Jun 30 00:00:00 EDT 2016},
month = {Thu Jun 30 00:00:00 EDT 2016}
}

Technical Report:

Save / Share:
  • Abstract not provided.
  • The purpose of this Individual Software Plan (ISP) is to plan and organize the activities required to certify the NUFT code for quality affecting work involving problems that include cross drift analysis of the Yucca Mountain Repository facility. NUFT is software for application to the solution of a class of coupled mass and heat transport problems in porous geologic media including Yucca Mountain Repository Cross Drift Problem (YMRCDP- also known as the Enhanced Characterization of the Repository Block (ECRB)). Solution of this class of problems requires a suite of multiphase, multi-component models for numerical solution of non-isothermal flow and transportmore » in porous media with applications to subsurface contaminant transport problems.« less
  • This document is the Individual Software Plan (ISP) for version 2.0 of the Non-isothermal Unsaturated-saturated Flow and Transport (NUFT.) analysis computer program. This document addresses the applicable requirements of LLNL YMP procedure 033-YMP-QP 3.2, Section 4.2.1.1. The purpose of this ISP is to plan and organize the activities required to certify the NUFT code for quality affecting work involving problems that include cross drift analysis of the Yucca Mountain Repository facility. NUFT is software for application to the solution of a class of coupled mass and heat transport problems in porous geologic media including Yucca Mountain Repository Cross Drift Problemmore » (YMRCDP- also known as the Enhanced Characterization of the Repository Block (ECRB)). Solution of this class of problems requires a suite of multiphase, multi-component models for numerical solution of non- isothermal flow and transport in porous media with applications to subsurface contaminant transport problems. NUFT is a suite of multiphase, multi-component models for numerical solution of non- isothermal flow and transport in porous media, with application to subsurface contaminant transport problems, and in particular, to the hydrology in and about the Yucca Mountain Repository Site. NUFI is acquired software, as defined by 033-YMP-QP 3.2, and a preliminary baseline of source code, electronic documentation, and paper documentation has been established as required by 033-YMP-QP 3.2, Section 4.1. NUFT runs on Sun Unix platforms, Solaris operating system version 5.5 and HP-UX with operating system version 10.20. The product to be qualified under this ISP is the version running on HP- UX. The HP version will be labeled Version 2.0h. The "h" is included to distinguish the HP version from possible future versions qualified for Sun or other platforms. The scope of the plans and procedures outlined in this ISP is limited to the effort required to qualify NUFT for the class of problems identified in section 1.1 above, to meet the applicableiequirements of 033-YMP-QP 3.2, and to establish a Version 2.0 baseline for subsequent qualifications of NUFT for additional problems or for use on additional platforms. The detailed requirements applicable to the class of problems for which this release is applicable will be defined in the NUFT 2.0 Requirements Document. The scope of this ISP is limited to NUFT. Auxiliary software such as the post-processing tool XTOOL, YMESH, -and RADPRO, which are often used in conjunction with NUFT, will be addressed separately as needed. Version 2.0 is the first quality controlled release for the NUFT software. Additional releases are anticipated. NUFT Version 2.0 is "acquired software," as defined by 033- YMP-QP 3.2. Version 1 .O was released in February 1998, but that certification is being withdrawn based upon the quality assurance surveillance findings described in LLNL-98- D-065.« less
  • All of the work conducted under the auspices of DE-FC02-01ER25473 was characterized by exceptionally close collaboration with researchers at the Lawrence Berkeley National Laboratory (LBNL). This included having one of my graduate students - Sarah Williams - spend the summer working with Dr. Ann Almgren a staff scientist in the Center for Computational Sciences and Engineering (CCSE) which is a part of the National Energy Research Supercomputer Center (NERSC) at LBNL. As a result of this visit Sarah decided to work on a problem suggested by Dr. John Bell the head of CCSE for her PhD thesis, which she finishedmore » in June 2007. Writing a PhD thesis while working at one of the University of California (UC) managed DOE laboratories is a long established tradition at the University of California and I have always encouraged my students to consider doing this. For example, in 2000 one of my graduate students - Matthew Williams - finished his PhD thesis while working with Dr. Douglas Kothe at the Los Alamos National Laboratory (LANL). Matt is now a staff scientist in the Diagnostic Applications Group in the Applied Physics Division at LANL. Another one of my graduate students - Christopher Algieri - who was partially supported with funds from DE-FC02-01ER25473 wrote am MS Thesis that analyzed and extended work published by Dr. Phil Colella and his colleagues in 1998. Dr. Colella is the head of the Applied Numerical Algorithms Group (ANAG) in the National Energy Research Supercomputer Center at LBNL and is the lead PI for the APDEC ISIC which was comprised of several National Laboratory research groups and at least five University PI's at five different universities. Chris Algieri is now employed as a staff member in Dr. Bill Collins' research group at LBNL developing computational models for climate change research. Bill Collins was recently hired at LBNL to start and be the Head of the Climate Science Department in the Earth Sciences Division at LBNL. Prior to this he had been a Deputy Section Head at the National Center for Atmospheric Research in Colorado. My understanding is that Chris Algieri is the first person that Bill hired after coming to LBNL. The plan is that Chris Algieri will finish his PhD thesis while employed as a staff scientist in Bill's group. Both Sarah and Chris were supported in part with funds from DE-FC02-01ER25473. In Sarah's case she received support both while at U.C. Davis (UCD) taking classes and writing an MS thesis and during some of the time she was living in Berkeley, working at LBNL and finishing her PhD thesis. In Chris' case he was at U.C. Davis during the entire time he received support from DE-FC02-01ER25473. More specific details of their work are included in the report below. Finally my own research conducted under the auspices of DE-FC02-01ER25473 either involved direct collaboration with researchers at LBNL - Phil Colella and Peter Schwartz who is a member of Phil's Applied Numerical Algorithms Group - or was on problems that are closely related to research that has been and continues to be conducted by researchers at LBNL. Specific details of this work can be found below. Finally, I would like to note that the work conducted by my students and me under the auspices of this contract is closely related to work that I have performed with funding from my DOE MICS contract DE-FC02-03ER25579 'Development of High-Order Accurate Interface Tracking Algorithms and Improved Constitutive Models for Problems in Continuum Mechanics with Applications to Jetting' and with my CoPI on that grant Professor Greg Miller of the Department of Applied Science at UCD. In theory I tried to use funds from the SciDAC grant DE-FC02-01ER25473 to support work that directly involved implementing algorithms developed by my research group at U.C. Davis in software that was developed and is maintained by my SciDAC CoPI's at LBNL.« less
  • The volume examines monitoring and reporting procedures applicable to performance and productivity. The managers of three transit systems of varying sizes discuss how productivity has been increased within their systems. The Pittsburgh discussion reviews a variety of capital projects and management improvement initiatives undertaken by Port Authority Transit, including development of maintenance manuals and an absenteeism reduction project. The Houston discussion describes how the property used a management study to identify needed organizational change and restructuring. The Ann Arbor discussion focuses on a set of service standards and performance indicators adopted to focus the management improvement process.