SpringerLink
Log in

Data Transfer and Reuse Analysis Tool for GPU-Offloading Using OpenMP

  • Conference paper
  • First Online:
OpenMP: Portable Multi-Level Parallelism on Modern Systems (IWOMP 2020)

Part of the book series: Lecture Notes in Computer Science ((LNPSE,volume 12295))

Included in the following conference series:

Abstract

In the high performance computing sector, researchers and application developers expend considerable effort to port their applications to GPU-based clusters in order to take advantage of the massive parallelism and energy efficiency of a GPU. Unfortunately porting or writing an application for accelerators, such as GPUs, requires extensive knowledge of the underlying architectures, the application/algorithm and the interfacing programming model, such as CUDA, HIP or OpenMP. Compared to native GPU programming models, OpenMP has a shorter learning curve, is portable and potentially also performance portable. To reduce the developer effort, OpenMP provides implicit data transfer between CPU and GPU. OpenMP users may control the duration of a data object’s allocation on the GPU via the use of target data regions, but they do not need to. Unfortunately, unless data map**s are explicitly provided by the user, compilers like Clang move all data accessed by a kernel to the GPU without considering its prior availability on the device. As a result, applications may spend a significant portion of their execution time on data transfer. Yet exploiting data reuse opportunities in an application has the potential to significantly reduce the overall execution time. In this paper we present a source-to-source tool that automatically identifies data in an OpenMP program which do not need to be transferred between CPU and GPU. The tool capitalizes on any data reuse opportunities to insert the pertinent, optimized OpenMP target data directives. Our experimental results show considerable reduction in the overall execution time of a set of micro-benchmarks and some benchmark applications from the Rodinia benchmark suite. To the best of our knowledge, no other tool optimizes OpenMP data map**s by identifying and exploiting data reuse opportunities between kernels.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Log in via an institution

Subscribe and save

Springer+ Basic
EUR 32.99 /Month
  • Get 10 units per month
  • Download Article/Chapter or Ebook
  • 1 Unit = 1 Article or 1 Chapter
  • Cancel anytime
Subscribe now

Buy Now

Chapter
EUR 29.95
Price includes VAT (Thailand)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
EUR 64.19
Price includes VAT (Thailand)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Softcover Book
EUR 74.99
Price excludes VAT (Thailand)
  • Compact, lightweight edition
  • Dispatched in 3 to 5 business days
  • Free ship** worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Similar content being viewed by others

Notes

  1. 1.

    In this paper the term kernel is always used in reference to a GPU kernel.

References

  1. Barua, P., Shirako, J., Tsang, W., Paudel, J., Chen, W., Sarkar, V.: OMPSan: static verification of OpenMP’s data map** constructs. In: Fan, X., de Supinski, B.R., Sinnen, O., Giacaman, N. (eds.) IWOMP 2019. LNCS, vol. 11718, pp. 3–18. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-28596-8_1

    Chapter  Google Scholar 

  2. Bercea, G.T., et al.: Implementing implicit OpenMP data sharing on GPUs. In: Proceedings of the Fourth Workshop on the LLVM Compiler Infrastructure in HPC, pp. 1–12 (2017)

    Google Scholar 

  3. C++ Heterogeneous-Compute Interface for Portability (2016). https://github.com/ROCm-Developer-Tools/HIP

  4. Che, S., et al.: Rodinia: a benchmark suite for heterogeneous computing. In: 2009 IEEE International Symposium on Workload Characterization (IISWC), pp. 44–54. IEEE (2009)

    Google Scholar 

  5. Clang 8.0 (2019). http://releases.llvm.org/8.0.1/tools/clang/docs/index.html

  6. Clang, Libtooling (2019). http://clang.llvm.org/docs/LibTooling.html

  7. Consortium, O., et al.: OpenMP specification version 5.0 (2018)

    Google Scholar 

  8. Cray, C.: C++ reference manual, s-2179 (8.7). Cray Research (2019). https://pubs.cray.com/content/S-2179/8.7/cray-c-and-c++-reference-manual/openmp-overview

  9. Dagum, L., Menon, R.: OpenMP: an industry standard API for shared-memory programming. IEEE Comput. Sci. Eng. 5(1), 46–55 (1998)

    Article  Google Scholar 

  10. Dulloor, S.R., et al.: Data tiering in heterogeneous memory systems. In: Proceedings of the Eleventh European Conference on Computer Systems, pp. 1–16 (2016)

    Google Scholar 

  11. Garcia, V., Debreuve, E., Barlaud, M.: Fast k nearest neighbor search using GPU. In: 2008 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, pp. 1–6. IEEE (2008)

    Google Scholar 

  12. GCC Support for the OpenMP Language (2019). https://gcc.gnu.org/wiki/openmp

  13. Gelado, I., Stone, J.E., Cabezas, J., Patel, S., Navarro, N., Hwu, W.M.W.: An asymmetric distributed shared memory model for heterogeneous parallel systems. In: Proceedings of the Fifteenth International Conference on Architectural Support for Programming Languages and Operating Systems, pp. 347–358 (2010)

    Google Scholar 

  14. Goodrum, M.A., Trotter, M.J., Aksel, A., Acton, S.T., Skadron, K.: Parallelization of particle filter algorithms. In: Varbanescu, A.L., Molnos, A., van Nieuwpoort, R. (eds.) ISCA 2010. LNCS, vol. 6161, pp. 139–149. Springer, Heidelberg (2011). https://doi.org/10.1007/978-3-642-24322-6_12

    Chapter  Google Scholar 

  15. Harish, P., Narayanan, P.J.: Accelerating large graph algorithms on the GPU using CUDA. In: Aluru, S., Parashar, M., Badrinath, R., Prasanna, V.K. (eds.) HiPC 2007. LNCS, vol. 4873, pp. 197–208. Springer, Heidelberg (2007). https://doi.org/10.1007/978-3-540-77220-0_21

    Chapter  Google Scholar 

  16. Huang, W., Ghosh, S., Velusamy, S., Sankaranarayanan, K., Skadron, K., Stan, M.R.: Hotspot: a compact thermal modeling methodology for early-stage VLSI design. IEEE Trans. Very Large Scale Integr. (VLSI) Syst. 14(5), 501–513 (2006)

    Article  Google Scholar 

  17. Intel C++ Compiler Code Samples (March 2019). https://software.intel.com/en-us/code-samples/intel-c-compiler

  18. Jablin, T.B., Prabhu, P., Jablin, J.A., Johnson, N.P., Beard, S.R., August, D.I.: Automatic CPU-GPU communication management and optimization. In: Proceedings of the 32nd ACM SIGPLAN Conference on Programming Language Design and Implementation, pp. 142–151 (2011)

    Google Scholar 

  19. Lattner, C., Adve, V.: LLVM: a compilation framework for lifelong program analysis & transformation. In: Proceedings of the International Symposium on Code Generation and Optimization: Feedback-Directed and Runtime Optimization, p. 75. IEEE Computer Society (2004)

    Google Scholar 

  20. Li, L., Chapman, B.: Compiler assisted hybrid implicit and explicit GPU memory management under unified address space. In: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, pp. 1–16 (2019)

    Google Scholar 

  21. LLVM Support for the OpenMP Language (2019). https://openmp.llvm.org

  22. Mendonça, G., Guimarães, B., Alves, P., Pereira, M., Araújo, G., Pereira, F.M.Q.: DawnCC: automatic annotation for data parallelism and offloading. ACM Trans. Archit. Code Optim. (TACO) 14(2), 13 (2017)

    Google Scholar 

  23. Mishra, A., Kong, M., Chapman, B.: Kernel fusion/decomposition for automatic GPU-offloading. In: Proceedings of the 2019 IEEE/ACM International Symposium on Code Generation and Optimization, pp. 283–284. IEEE Press (2019)

    Google Scholar 

  24. Mishra, A., Li, L., Kong, M., Finkel, H., Chapman, B.: Benchmarking and evaluating unified memory for OpenMP GPU offloading. In: Proceedings of the Fourth Workshop on the LLVM Compiler Infrastructure in HPC, pp. 1–10 (2017)

    Google Scholar 

  25. Nvidia, C.: Nvidia cuda c programming guide. Nvidia Corp. 120(18), 8 (2011)

    Google Scholar 

  26. NVIDIA Tesla: Nvidia tesla v100 GPU architecture (2017)

    Google Scholar 

  27. OpenMP Compilers & Tools (April 2019). https://www.openmp.org/resources/openmp-compilers-tools

  28. Poesia, G., Guimarães, B., Ferracioli, F., Pereira, F.M.Q.: Static placement of computation on heterogeneous devices. Proc. ACM Program. Lang. 1(OOPSLA), 50 (2017)

    Article  Google Scholar 

  29. Poesia, G., Guimarães, B.C.F., Ferracioli, F., Pereira, F.M.Q.: Static placement of computation on heterogeneous devices. Proc. ACM Program. Lang. 1(OOPSLA), 50:1–50:28 (2017). Article 50

    Google Scholar 

  30. Seawulf, Computational Cluster at Stony Brook University (2019). https://it.stonybrook.edu/help/kb/understanding-seawulf

  31. Vazhkudai, S.S., et al.: The design, deployment, and evaluation of the coral pre-exascale systems. In: SC18: International Conference for High Performance Computing, Networking, Storage and Analysis, pp. 661–672. IEEE (2018)

    Google Scholar 

  32. Wienke, S., Springer, P., Terboven, C., an Mey, D.: OpenACC—first experiences with real-world applications. In: Kaklamanis, C., Papatheodorou, T., Spirakis, P.G. (eds.) Euro-Par 2012. LNCS, vol. 7484, pp. 859–870. Springer, Heidelberg (2012). https://doi.org/10.1007/978-3-642-32820-6_85

    Chapter  Google Scholar 

  33. Yu, S., Park, S., Baek, W.: Design and implementation of bandwidth-aware memory placement and migration policies for heterogeneous memory systems. In: Proceedings of the International Conference on Supercomputing, pp. 1–10 (2017)

    Google Scholar 

Download references

Acknowledgement

This research was supported by the Exascale Computing Project (17-SC-20-SC), a collaborative effort of the U.S. Department of Energy Office of Science and the National Nuclear Security Administration. The authors would like to thank Stony Brook Research Computing and Cyberinfrastructure, and the Institute for Advanced Computational Science at Stony Brook University for access to the SeaWulf computing system, which was made possible by a $1.4M National Science Foundation grant (#1531492). Special thanks to our colleague Dr. Chunhua Liao from Lawrence Livermore National Laboratory for his initial feedback and helpful discussions.

Author information

Authors and Affiliations

  1. Stony Brook University, Stony Brook, NY, 11794, USA

    Alok Mishra & Barbara Chapman

  2. Brookhaven National Laboratory, Upton, NY, 11973, USA

    Abid M. Malik & Barbara Chapman

Authors
  1. Alok Mishra
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Abid M. Malik
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Barbara Chapman
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Alok Mishra .

Editor information

Editors and Affiliations

  1. Texas Advanced Computing Center (TACC), Austin, TX, USA

    Kent Milfeld

  2. Lawrence Livermore National Laboratory, Livermore, CA, USA

    Bronis R. de Supinski

  3. Texas Advanced Computing Center (TACC), Austin, TX, USA

    Lars Koesterke

  4. RWTH Aachen University, Aachen, Germany

    Jannis Klinkenberg

Rights and permissions

Reprints and permissions

Copyright information

© 2020 Springer Nature Switzerland AG

About this paper

Check for updates. Verify currency and authenticity via CrossMark

Cite this paper

Mishra, A., Malik, A.M., Chapman, B. (2020). Data Transfer and Reuse Analysis Tool for GPU-Offloading Using OpenMP. In: Milfeld, K., de Supinski, B., Koesterke, L., Klinkenberg, J. (eds) OpenMP: Portable Multi-Level Parallelism on Modern Systems. IWOMP 2020. Lecture Notes in Computer Science(), vol 12295. Springer, Cham. https://doi.org/10.1007/978-3-030-58144-2_18

Download citation

  • DOI: https://doi.org/10.1007/978-3-030-58144-2_18

  • Published:

  • Publisher Name: Springer, Cham

  • Print ISBN: 978-3-030-58143-5

  • Online ISBN: 978-3-030-58144-2

  • eBook Packages: Computer ScienceComputer Science (R0)

Publish with us

Policies and ethics

Search

Navigation