Abstract
Computer vision applications generate and process a huge amount of data with the help of heterogeneous high-performance computing to achieve or attain human-level accuracy. One of the challenging tasks is mapping the required data across the memory hierarchy in the heterogeneous environment effectively. Irregular memory access and thread divergence cause bottleneck in the interconnect data communication. In this paper, a thread geometry analysis is proposed to perform permutation through the primary axis and calculate the cost of each axis based on the array references made. The permutation, which produces the lowest cost is considered to have a least uncoalesced access among the other axes. It reduces the global load transaction per request by 76% and stores transactions per request by 96%. The reduction of the number of Loads and Stores has improved the IPC by a factor of 2.3\(\times \) for NW benchmark.
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
Hwu, W.M.: What is ahead for parallel computing. J. Parallel Distrib. Comput. 74(7), 2574–2581 (2014)
Deshpande, P., Sharma, S.C., Peddoju, S.K., Abraham, A.: Efficient multimedia data storage in cloud environment. Informatica, 39(4), 2015
Tzeng, S., Patney, A., Owens, J.D.: Task management for irregular-parallel workloads on the GPU. In: Proceedings of the Conference on High Performance Graphics, pp. 29–37. Eurographics Association (2010)
Huo, X., Ravi, V.T., Agrawal, G.: Porting irregular reductions on heterogeneous CPU-GPU configurations. In: High Performance Computing (HiPC), 2011 18th International Conference on, pp. 1–10. IEEE (2011)
Liu, W., Vinter, B.: An efficient GPU general sparse matrix-matrix multiplication for irregular data. In: Parallel and Distributed Processing Symposium, 2014 IEEE 28th International, pp. 370–381. IEEE (2014)
Zhang, J., Wang, H., Feng, W.C.: Cublastp: fine-grained parallelization of protein sequence search on CPU+GPU. IEEE/ACM Trans. Comput. Biol. Bioinform. (TCBB) 14(4), 830–843 (2017)
Caragea, G.C., Keceli, F., Tzannes, A., Vishkin, U.: General-purpose versus GPU: comparison of many-cores on irregular workloads. In: HotPar’10: Proceedings of the 2nd Workshop on Hot Topics in Parallelism (2010)
Burtscher, M., Nasre, R., Pingali, K.: A quantitative study of irregular programs on GPUs. In: 2012 IEEE International Symposium on Workload Characterization (IISWC), pp. 141–151. IEEE (2012)
O’Neil, M.A, Burtscher, M.: Microarchitectural performance characterization of irregular GPU kernels. In: 2014 IEEE International Symposium on Workload Characterization (IISWC), pp. 130–139. IEEE (2014)
Ben-Nun, T., Sutton, M., Pai, S., Pingali, K.: Groute: an asynchronous multi-GPU programming model for irregular computations. In: ACM SIGPLAN Notices, vol. 52, pp. 235–248. ACM (2017)
Alur, R., Devietti, J., Leija, O.S.N., Singhania, N.: Gpudrano: detecting uncoalesced accesses in GPU programs. In: International Conference on Computer Aided Verification, pp. 507–525. Springer, Berlin (2017)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Singapore Pte Ltd.
About this paper
Cite this paper
Tamizharasan, P.S., Ramasubramanian, N. (2020). Enhancing GPU Performance Using Thread Geometry Analysis for Irregular Workloads. In: Iyer, B., Deshpande, P., Sharma, S., Shiurkar, U. (eds) Computing in Engineering and Technology. Advances in Intelligent Systems and Computing, vol 1025. Springer, Singapore. https://doi.org/10.1007/978-981-32-9515-5_19
Download citation
DOI: https://doi.org/10.1007/978-981-32-9515-5_19
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-32-9514-8
Online ISBN: 978-981-32-9515-5
eBook Packages: Intelligent Technologies and RoboticsIntelligent Technologies and Robotics (R0)