Accelerated application development: The ORNL Titan experience

Wayne Joubert; Rick Archibald; Mark Berrill; W. Michael Brown; Markus Eisenbach; Ray Grout; Jeff Larkin; John Levesque; Bronson Messer; Matt Norman; Bobby Philip; Ramanan Sankaran; Arnold Tharrington; John Turner

doi:10.1016/j.compeleceng.2015.04.008

Accelerated application development: The ORNL Titan experience

Wayne Joubert, Rick Archibald, Mark Berrill, W. Michael Brown, Markus Eisenbach, Ray Grout, Jeff Larkin, John Levesque, Bronson Messer, Matt Norman, Bobby Philip, Ramanan Sankaran, Arnold Tharrington, John Turner

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

26 Scopus citations

Abstract

The use of computational accelerators such as NVIDIA GPUs and Intel Xeon Phi processors is now widespread in the high performance computing community, with many applications delivering impressive performance gains. However, programming these systems for high performance, performance portability and software maintainability has been a challenge. In this paper we discuss experiences porting applications to the Titan system. Titan, which began planning in 2009 and was deployed for general use in 2013, was the first multi-petaflop system based on accelerator hardware. To ready applications for accelerated computing, a preparedness effort was undertaken prior to delivery of Titan. In this paper we report experiences and lessons learned from this process and describe how users are currently making use of computational accelerators on Titan.

Original language	English
Pages (from-to)	123-138
Number of pages	16
Journal	Computers and Electrical Engineering
Volume	46
DOIs	https://doi.org/10.1016/j.compeleceng.2015.04.008
State	Published - 2015

Keywords

Accelerated computing
Code refactoring
GPU graphics processing units
High performance computing
Science applications
Software optimization

Access to Document

10.1016/j.compeleceng.2015.04.008

Cite this

@article{a65fe20935c848f6866a10a23e365abd,

title = "Accelerated application development: The ORNL Titan experience",

abstract = "The use of computational accelerators such as NVIDIA GPUs and Intel Xeon Phi processors is now widespread in the high performance computing community, with many applications delivering impressive performance gains. However, programming these systems for high performance, performance portability and software maintainability has been a challenge. In this paper we discuss experiences porting applications to the Titan system. Titan, which began planning in 2009 and was deployed for general use in 2013, was the first multi-petaflop system based on accelerator hardware. To ready applications for accelerated computing, a preparedness effort was undertaken prior to delivery of Titan. In this paper we report experiences and lessons learned from this process and describe how users are currently making use of computational accelerators on Titan.",

keywords = "Accelerated computing, Code refactoring, GPU graphics processing units, High performance computing, Science applications, Software optimization",

author = "Wayne Joubert and Rick Archibald and Mark Berrill and {Michael Brown}, W. and Markus Eisenbach and Ray Grout and Jeff Larkin and John Levesque and Bronson Messer and Matt Norman and Bobby Philip and Ramanan Sankaran and Arnold Tharrington and John Turner",

year = "2015",

doi = "10.1016/j.compeleceng.2015.04.008",

language = "English",

volume = "46",

pages = "123--138",

journal = "Computers and Electrical Engineering",

issn = "0045-7906",

publisher = "Elsevier",

}

TY - JOUR

T1 - Accelerated application development

T2 - The ORNL Titan experience

AU - Joubert, Wayne

AU - Archibald, Rick

AU - Berrill, Mark

AU - Michael Brown, W.

AU - Eisenbach, Markus

AU - Grout, Ray

AU - Larkin, Jeff

AU - Levesque, John

AU - Messer, Bronson

AU - Norman, Matt

AU - Philip, Bobby

AU - Sankaran, Ramanan

AU - Tharrington, Arnold

AU - Turner, John

PY - 2015

Y1 - 2015

N2 - The use of computational accelerators such as NVIDIA GPUs and Intel Xeon Phi processors is now widespread in the high performance computing community, with many applications delivering impressive performance gains. However, programming these systems for high performance, performance portability and software maintainability has been a challenge. In this paper we discuss experiences porting applications to the Titan system. Titan, which began planning in 2009 and was deployed for general use in 2013, was the first multi-petaflop system based on accelerator hardware. To ready applications for accelerated computing, a preparedness effort was undertaken prior to delivery of Titan. In this paper we report experiences and lessons learned from this process and describe how users are currently making use of computational accelerators on Titan.

AB - The use of computational accelerators such as NVIDIA GPUs and Intel Xeon Phi processors is now widespread in the high performance computing community, with many applications delivering impressive performance gains. However, programming these systems for high performance, performance portability and software maintainability has been a challenge. In this paper we discuss experiences porting applications to the Titan system. Titan, which began planning in 2009 and was deployed for general use in 2013, was the first multi-petaflop system based on accelerator hardware. To ready applications for accelerated computing, a preparedness effort was undertaken prior to delivery of Titan. In this paper we report experiences and lessons learned from this process and describe how users are currently making use of computational accelerators on Titan.

KW - Accelerated computing

KW - Code refactoring

KW - GPU graphics processing units

KW - High performance computing

KW - Science applications

KW - Software optimization

UR - http://www.scopus.com/inward/record.url?scp=84928923259&partnerID=8YFLogxK

U2 - 10.1016/j.compeleceng.2015.04.008

DO - 10.1016/j.compeleceng.2015.04.008

M3 - Article

AN - SCOPUS:84928923259

SN - 0045-7906

VL - 46

SP - 123

EP - 138

JO - Computers and Electrical Engineering

JF - Computers and Electrical Engineering

ER -

Accelerated application development: The ORNL Titan experience

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this