Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

GPU Programming Productivity in Different Abstraction Paradigms

GPU Programming Productivity in Different Abstraction Paradigms Coprocessor architectures in High Performance Computing are prevalent in today’s scientific computing clusters and require specialized knowledge for proper utilization. Various alternative paradigms for parallel and offload computation exist, but little is known about the human factors impacts of using the different paradigms. With computer science student participants from the University of Nevada, Las Vegas with no previous exposure to Graphics Processing Unit programming, our study compared NVIDIA CUDA C/C++ as a control group and the Thrust library. The designers of Thrust claim their higher level of abstraction enhances programmer productivity. The trial was conducted on 91 participants and was administered through our computerized testing platform. Although the study was narrowly focused on the basic steps of an offloaded computation problem and was not intended to be a comprehensive evaluation of the superiority of one approach or the other, we found evidence that although Thrust was designed for ease of use, the abstractions tended to be confusing to students and in several cases diminished productivity. Specifically, abstractions in Thrust for (i) memory allocation through a C++ Standard Template Library-style vector library call, (ii) memory transfers between the host and Graphics Processing Unit coprocessor through an overloaded assignment operator, and (iii) execution of an offloaded routine through a generic transform library call instead of a CUDA kernel routine all performed either equal to or worse than CUDA. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png ACM Transactions on Computing Education (TOCE) Association for Computing Machinery

GPU Programming Productivity in Different Abstraction Paradigms

Download PDF
27 pages

Loading next page...
 
/lp/association-for-computing-machinery/gpu-programming-productivity-in-different-abstraction-paradigms-fRzMP7F8mp
Publisher
Association for Computing Machinery
Copyright
Copyright © 2020 ACM
ISSN
1946-6226
eISSN
1946-6226
DOI
10.1145/3418301
Publisher site
See Article on Publisher Site

Abstract

Coprocessor architectures in High Performance Computing are prevalent in today’s scientific computing clusters and require specialized knowledge for proper utilization. Various alternative paradigms for parallel and offload computation exist, but little is known about the human factors impacts of using the different paradigms. With computer science student participants from the University of Nevada, Las Vegas with no previous exposure to Graphics Processing Unit programming, our study compared NVIDIA CUDA C/C++ as a control group and the Thrust library. The designers of Thrust claim their higher level of abstraction enhances programmer productivity. The trial was conducted on 91 participants and was administered through our computerized testing platform. Although the study was narrowly focused on the basic steps of an offloaded computation problem and was not intended to be a comprehensive evaluation of the superiority of one approach or the other, we found evidence that although Thrust was designed for ease of use, the abstractions tended to be confusing to students and in several cases diminished productivity. Specifically, abstractions in Thrust for (i) memory allocation through a C++ Standard Template Library-style vector library call, (ii) memory transfers between the host and Graphics Processing Unit coprocessor through an overloaded assignment operator, and (iii) execution of an offloaded routine through a generic transform library call instead of a CUDA kernel routine all performed either equal to or worse than CUDA.

Journal

ACM Transactions on Computing Education (TOCE) – Association for Computing Machinery

Published: Oct 13, 2020

Keywords: Computer science education

References

  • High Performance Computing
    Devices, Advanced Micro
  • The Landscape of Parallel Computing Research: A View from Berkeley
    Asanovic, Krste; Bodik, Ras; Catanzaro, Bryan Christopher; Gebis, Joseph James; Husbands, Parry; Keutzer, Kurt; Patterson, David A.; et al,
  • OpenMP
    Board, OpenMP Architecture Review
  • High level data structures for GPGPU programming in a statically typed language
    Bourgoin, Mathias; Chailloux, Emmanuel; Lamotte, Jean-Luc
  • Software transactional memory vs
    Castor, Fernando; Oliveira, João Paulo; Santos, André L. M.
  • A course-based usability analysis of Cilk Plus and OpenMP
    Coblenz, Michael; Seacord, Robert; Myers, Brad; Sunshine, Joshua; Aldrich, Jonathan
  • Intel Cilk Plus
    Corporation, Intel
  • Intel Threading Building Blocks
    Corporation, Intel
  • Intel Xeon Phi Coprocessors
    Corporation, Intel
  • GPU vs CPU? What is GPU Computing
    Corporation, NVIDIA
  • 2017 (updated March 20, 2017)
    Corporation, NVIDIA
  • Smoking and carcinoma of the lung
    Doll, Richard; Hill, A. Bradford
  • Debugging concurrent software: A study using multithreaded sequence diagrams
    Fleming, Scott D.; Kraemer, Eileen; Stirewalt, R. E. Kurt; Dillon, Laura K.
  • A study of student strategies for the corrective maintenance of concurrent software
    Fleming, Scott D.; Kraemer, Eileen; Stirewalt, R. E. Kurt; Xie, Shaohua; Dillon, Laura K.
  • Quick Start Guide
    Gaburov, Evghenii
  • OpenCL: The Open Standard for Parallel Programming of Heterogeneous Systems
    Group, Khronos
  • An Even Easier Introduction to CUDA
    Harris, Mark
  • Thrust--Parallel Algorithms Library
    Hoberock, Jared; Bell, Nathan
  • Parallel programmer productivity: A case study of novice parallel programmers
    Hochstein, Lorin; Carver, Jeffrey; Shull, Forrest; Asgari, Sima; Basili, Victor; Hollingsworth, Jeffrey K.; Zelkowitz, Marvin V.
  • Evidence-Based Programming Language Design: A Philosophical and Methodological Exploration
    Kaijanaho, Antti-Juhani
  • GPU clusters for high-performance computing
    Kindratenko, Volodymyr V.; Enos, Jeremy J.; Shi, Guochun; Showerman, Michael T.; Arnold, Galen W.; Stone, John E.; Phillips, James C.; Hwu, Wen-Mei
  • Commonsense computing (episode 3): Concurrency and concert tickets
    Lewandowski, Gary; Bouvier, Dennis J.; McCartney, Robert; Sanders, Kate; Simon, Beth
  • Theoretical underpinnings of computing education research: What is the evidence? In Proceedings of the 10th Annual Conference on International Computing Education Research
    Malmi, Lauri; Sheard, Judy; Bednarik, Roman; Helminen, Juha; Kinnunen, Päivi; Korhonen, Ari; Myller, Niko
  • The CONSORT statement: Revised recommendations for improving the quality of reports of parallel-group randomised trials
    Moher, David; Schulz, Kenneth F.; Altman, Douglas G.
  • Empirical assessment of languages for teaching concurrency: Methodology and application
    Nanz, Sebastian; Torshizi, Faraz; Pedroni, Michela; Meyer, Bertrand
  • A study of transactional memory vs
    Pankratius, Victor; Adl-Tabatabai, Ali-Reza
  • Is transactional programming actually easier? ACM SIGPLAN Notices 45, 5 (2010), 47--56
    Rossbach, Christopher J.; Hofmann, Owen S.; Witchel, Emmett
  • Programmers’ build errors: A case study (at Google)
    Seo, Hyunmin; Sadowski, Caitlin; Elbaum, Sebastian; Aftandilian, Edward; Bowdidge, Robert
  • CUDA for Engineers: An Introduction to High-Performance Parallel Computing
    Storti, Duane; Yurtoglu, Mete
  • CUDA-Lite: Reducing GPU programming complexity
    Ueng, S.-Z.; Lathara, M.; Baghsorkhi, S. S.; Hwu, W.-M. W.
  • Programming Difficulty Is Killing Engineers’ Productivity
    World, Scientific Computing