Abstract
With the increasing diversity of heterogeneous architecture in the HPC industry, porting a legacy application to run on different architectures is a tough challenge. In this paper, we present OpenMP Advisor, a novel compiler tool that enables code offloading to a GPU with OpenMP using Machine Learning. Although the tool is currently limited to GPUs, it can be extended to support other OpenMP-capable devices. The tool has two modes: Training and Prediction. It analyzes benchmark codes, generates every possible code variant on the target device, runs and gathers data to train an ML-based cost model in the training mode, which predicts the runtime of every code variant in the prediction mode. The main objective behind this tool is to maintain the portability aspect of OpenMP. Our Advisor produced code for several applications on seven architectures with four compilers, and accurately anticipated the top ten options for each application on every architecture. Initial results suggest that this tool can help compiler developers and HPC researchers migrate their legacy codes to the new heterogeneous computing environment.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Barua, P., Shirako, J., Tsang, W., Paudel, J., Chen, W., Sarkar, V.: OMPSan: static verification of OpenMP’s data mapping 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
Brower, R., Christ, N., DeTar, C., Edwards, R., Mackenzie, P.: Lattice QCD application development within the us doe exascale computing project. EPJ Web Conf. 175, 09010 (2018). https://doi.org/10.1051/epjconf/201817509010
Burford, A., et al.: Ookami: deployment and initial experiences. In: Practice and Experience in Advanced Research Computing, pp. 1–8. ACM, New York (2021)
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)
Clang: Clang Rewriter class reference (2021). https://clang.llvm.org/doxygen/classclang_1_1Rewriter.html
Dagum, L., Menon, R.: OpenMP: an industry standard API for shared-memory programming. IEEE Comput. Sci. Eng. 5(1), 46–55 (1998)
Glorot, X., Bengio, Y.: Understanding the difficulty of training deep feedforward neural networks. In: Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, pp. 249–256. JMLR Workshop and Conference Proceedings (2010)
Hinton, G., Srivastava, N., Swersky, K.: Neural networks for machine learning. Lecture 6a Overview of Mini-batch Gradient Descent, vol. 14, no. 8, p. 2 (2012)
Intel: Intel Developer Cloud (2021). https://www.intel.com/content/www/us/en/developer/tools/devcloud/overview.html
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)
Kale, V., Lu, W., Curtis, A., Malik, A.M., Chapman, B., Hernandez, O.: Toward supporting multi-GPU targets via taskloop and user-defined schedules. In: Milfeld, K., de Supinski, B.R., Koesterke, L., Klinkenberg, J. (eds.) IWOMP 2020. LNCS, vol. 12295, pp. 295–309. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-58144-2_19
Kingma, D.P., Ba, J.: Adam: a method for stochastic optimization. In: Bengio, Y., LeCun, Y. (eds.) 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, 7–9 May 2015, Conference Track Proceedings (2015). http://arxiv.org/abs/1412.6980
Laboratory, L.L.N.: LLNL - Corona (2019). https://hpc.llnl.gov/hardware/compute-platforms/corona
Lattner, C., Adve, V.: LLVM: a compilation framework for lifelong program analysis & transformation. In: International Symposium on Code Generation and Optimization, 2004. CGO 2004, pp. 75–86. IEEE (2004)
Lin, M.: Optimization of the domain wall dslash kernel in columbia physics system, p. 269 (2016)
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 Optimiz. (TACO) 14(2), 13 (2017)
Mishra, A., Chheda, S., Soto, C., Malik, A.M., Lin, M., Chapman, B.: COMPOFF: a compiler cost model using machine learning to predict the cost of openmp offloading. In: 2022 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), 30 May - 3 June 2022. IEEE (2022)
Mishra, A., Malik, A.M., Chapman, B.: Data transfer and reuse analysis tool for GPU-offloading using openMP. In: Milfeld, K., de Supinski, B.R., Koesterke, L., Klinkenberg, J. (eds.) IWOMP 2020. LNCS, vol. 12295, pp. 280–294. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-58144-2_18
Mishra, A., Malik, A.M., Chapman, B.: Extending the LLVM/clang framework for openMP metadirective support. In: 2020 IEEE/ACM 6th Workshop on the LLVM Compiler Infrastructure in HPC (LLVM-HPC) and Workshop on Hierarchical Parallelism for Exascale Computing (HiPar), pp. 33–44. IEEE (2020)
ORNL: Oak Ridge Leadership Computing Facility - Summit supercomputing cluster (2017). https://www.olcf.ornl.gov/summit/
ORNL: Oak Ridge Leadership Computing Facility - Wombat cluster (2020). https://www.olcf.ornl.gov/olcf-resources/compute-systems/wombat/
Poesia, G., Guimarães, B., Ferracioli, F., Pereira, F.M.Q.: Static placement of computation on heterogeneous devices. In: Proceedings of the ACM on Programming Languages 1(OOPSLA), pp. 1–28 (2017)
Stony Brook University: Seawulf, computational cluster at stony brook university (2019). https://it.stonybrook.edu/help/kb/understanding-seawulf
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. This material is also based upon work supported by the National Science Foundation under grant no. CCF-2113996. This research used resources of the Oak Ridge Leadership Computing Facility at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725. 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).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Mishra, A., Malik, A.M., Lin, M., Chapman, B. (2023). OpenMP Advisor: A Compiler Tool for Heterogeneous Architectures. In: McIntosh-Smith, S., Klemm, M., de Supinski, B.R., Deakin, T., Klinkenberg, J. (eds) OpenMP: Advanced Task-Based, Device and Compiler Programming. IWOMP 2023. Lecture Notes in Computer Science, vol 14114. Springer, Cham. https://doi.org/10.1007/978-3-031-40744-4_3
Download citation
DOI: https://doi.org/10.1007/978-3-031-40744-4_3
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-40743-7
Online ISBN: 978-3-031-40744-4
eBook Packages: Computer ScienceComputer Science (R0)