Abstract
Heterogeneous processors, comprising CPU cores and a GPU, are the de facto standard in desktop and mobile platforms. In many cases it is worthwhile to exploit both the CPU and GPU simultaneously. However, the workload distribution poses a challenge when running irregular applications. In this paper, we present LogFit, a novel adaptive partitioning strategy for parallel loops, specially designed for applications with irregular data accesses running on heterogeneous CPU–GPU architectures. Our algorithm dynamically finds the optimal chunk size that must be assigned to the GPU. Also, the number of iterations assigned to the CPU cores are adaptively computed to avoid load unbalance. In addition, we also strive to increase the programmer’s productivity by providing a high level template that eases the coding of heterogeneous parallel loops. We evaluate LogFit’s performance and energy consumption by using a set of irregular benchmarks running on a heterogeneous CPU–GPU processor, an Intel Haswell. Our experimental results show that we outperform Oracle-like static and other dynamic state-of-the-art approaches both in terms of performance, up to 57%, and energy saving, up to 31%.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
For instance, Threading Building Blocks library (TBB) [22], recommends to have CPU chunk sizes that take 100,000 clock cycles at least.
RO = Read-Only; WO = Write-Only; RW = Read–Write
\(nEU =\,\)clGetDeviceInfo(deviceId, CL_DEVICE_MAX_COMPUTE_UNITS)
References
Augonnet, C., Clet-Ortega, J., Thibault, S., Namyst, R.: Data-aware task scheduling on multi-accelerator based platforms. In: Proceedings of ICPADS, pp. 291–298 (2010)
Belviranli, M., Bhuyan, L., Gupta, R.: A dynamic self-scheduling scheme for heterogeneous multiprocessor architectures. ACM Trans. Archit. Code Optim. 9(4), 57 (2013)
Bueno, J., Planas, J., Duran, A., Badia, R., Martorell, X., Ayguade, E., Labarta, J.: Productive programming of GPU clusters with OmpSs. In: Proceedings of IPDPS (2012)
Burtscher, M., Nasre, R., Pingali, K.: A quantitative study of irregular programs on GPUs. In: Proceedings of IISWC, pp. 141–151 (2012)
Chatterjee, S., Grossman, M., Sbirlea, A., Sarkar, V.: Dynamic task parallelism with a GPU work-stealing runtime system. In: LNCS Series, vol. 7146, pp. 203–217 (2011)
Che, S., et al.: A characterization of the Rodinia benchmark suite with comparison to contemporary CMP workloads. In: IISWC, pp. 1–11 (2010)
Danalis, A., Marin, G., McCurdy, C., et al.: The scalable heterogeneous computing (SHOC) benchmark suite. In: GPGPU, pp. 63–74 (2010)
Davis, T.A., Hu, Y.: The University of Florida sparse matrix collection. ACM Trans. Math. Softw. 38(1), 1–25 (2011)
Dementiev, R., Willhalm, T., Bruggeman, O., et al.: Intel Performance Counter Monitor (2012). www.intel.com/software/pcm
Gibbon, P., Frings, W., Mohr, B.: Performance analysis and visualization of the n-body tree code PEPC on massively parallel computers. In: PARCO, pp. 367–374 (2005)
Hart, A.: The OpenACC programming model. Technical Report, Cray Exascale Research Initiative Europe (2012)
Intel: Intel OpenCL N-Body Sample (2014)
Intel VTune Amplifier 2015 (2014). https://software.intel.com/en-us/intel-vtune-amplifier-xe
Kaleem, R., et al.: Adaptive heterogeneous scheduling for integrated GPUs. In: International Conference on Parallel Architectures and Compilation, PACT ’14, pp. 151–162 (2014)
Kulkarni, M., Burtscher, M., Cascaval, C., Pingali, K.: Lonestar: a suite of parallel irregular programs. In: ISPASS, pp. 65–76 (2009)
Li, D., Rhu, M., et al.: Priority-based cache allocation in throughput processors. In: International Symposium on High Performance Computer Architecture (HPCA) (2015)
Lima, J., Gautier, T., Maillard, N., Danjean, V.: Exploiting concurrent GPU operations for efficient work stealing on multi-GPUs. In: SBAC-PAD’12, pp. 75–82 (2012)
Luk, C.K., Hong, S., Kim, H.: Qilin: exploiting parallelism on heterogeneous multiprocessors with adaptive mapping. In: Proceedings of Microarchitecture, pp. 45–55 (2009)
Navarro, A., Vilches, A., Corbera, F., Asenjo, R.: Strategies for maximizing utilization on multi-CPU and multi-GPU heterogeneous architectures. J. Supercomput. 70, 756–771 (2014)
NVidia: CUDA Toolkit 5.0 Performance Report (2013)
Pandit, P., Govindarajan, R.: Fluidic kernels: cooperative execution of OpenCL programs on multiple heterogeneous devices. In: CGO (2014)
Reinders, J.: Intel Threading Building Blocks: Outfitting C++ for Multi-core Processor Parallelism. O’Reilly Media, Inc. (2007)
Rogers, T.G., O’Connor, M., Aamodt, T.M.: Cache-conscious wavefront scheduling. In: IEEE/ACM International Symposium on Microarchitecture, MICRO-45 (2012)
Russel, S.: Levering GPGPU and OpenCL technologies for natural user interaces. Technical Report, You i Labs inc (2012)
Sbirlea, A., Zou, Y., Budimlic, Z., Cong, J., Sarkar, V.: Mapping a data-flow programming model onto heterogeneous platforms. In: Proceedings of LCTES, pp. 61–70 (2012)
Wang, Z., Zheng, L., Chen, Q., Guo, M.: CPU + GPU scheduling with asymptotic profiling. Parallel Comput. 40(2), 107–115 (2014)
Acknowledgements
Funding was provided by Ministerio de Economía y Competitividad (Grant No. TIN2016-80920-R) and Consejería de Economía, Innovación, Ciencia y Empleo, Junta de Andalucía P11-(Grant No. TIC-08144).
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Navarro, A., Corbera, F., Rodriguez, A. et al. Heterogeneous parallel_for Template for CPU–GPU Chips. Int J Parallel Prog 47, 213–233 (2019). https://doi.org/10.1007/s10766-018-0555-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10766-018-0555-0