High Performance Computing on the IBM Power8 Platform
This paper discusses the performance of IBM’s Power8 CPUs, on a number of skeleton, financial and CFD benchmarks and applications. Implicitly, the performance of the software toolchain is also tested - the bare-bones Little-Endian Ubuntu, the GNU 5.3 and the XL 14.1.3 compilers and OpenMP runtimes. First, we aim to establish some roofline numbers on bandwidth and compute throughput, then move on to benchmark explicit and implicit one-/three-factor Black-Scholes computations, and CFD applications based on the OP2 and OPS frameworks, such as the Airfoil and BookLeaf unstructured-mesh codes, and the CloverLeaf 2D/3D structured mesh simulations. These applications all exhibit different characteristics in terms of computations, communications, memory access patterns, etc. Finally we briefly discuss performance of an industrial CFD code, Rolls-Royce Hydra, and we show initial results from IBM’s CUDA Fortran compiler. Both absolute and relative performance metrics are computed and compared to NVIDIA GPUs and Intel Xeon CPUs.
KeywordsDouble Precision Single Precision Memory Access Pattern POWER8 Platform Stencil Computation
The OP2 project has been funded by the UK Technology Strategy Board and Rolls-Royce plc. through the Siloet project, the UK Engineering and Physical Sciences Research Council projects
EP/I006079/1, EP/I00677X/1 on “Multi-layered Abstractions for PDEs”.
The OPS project is funded by the UK Engineering and Physical Sciences Research Council projects EP/K038494/1, EP/K038486/1, EP/K038451/1 and EP/K038567/1 on “Future-proof massively-parallel execution of multi-block applications” and EP/J010553/1 “Software for Emerging Architectures” (ASEArch) project. Cloverleaf development is supported by the Royal Society through their Industry Fellowship Scheme (IF090020/AM).
The authors would like to thank Chris Bowler, Carlo Bertolli, Wang Chen and Kelvin Li at IBM for their help in accessing expertise and software for the development of these experiments.
- 1.Bookleaf mini-app. https://github.com/UK-MAC/BookLeaf_ref. Accessed 4 Jan 2016
- 2.Adinetz, A.V., Baumeister, P.F., Böttiger, H., Hater, T., Maurer, T., Pleiter, D., Schenck, W., Schifano, S.F.: Performance evaluation of scientific applications on POWER8. In: Jarvis, S.A., Wright, S.A., Hammond, S.D. (eds.) PMBS 2014. LNCS, vol. 8966, pp. 24–45. Springer, Heidelberg (2015). doi: 10.1007/978-3-319-17248-4_2 Google Scholar
- 3.Giles, M., László, E., Reguly, I., Appleyard, J., Demouth, J.: GPU implementation of finite difference solvers. In: Proceedings of the 7th Workshop on High Performance Computational Finance, pp. 1–8. IEEE Press (2014)Google Scholar
- 4.Heroux, M., Barrett, R.: Mantevo project (2011)Google Scholar
- 5.McCalpin, J.D.: Memory bandwidth and machine balance in current high performance computers. In: IEEE Computer Society Technical Committee on Computer Architecture (TCCA) Newsletter, pp. 19–25, December 1995Google Scholar
- 6.Reguly, I.Z., Keita, A.-K., Giles, M.B.: Benchmarking the ibm power8 processor. In: Proceedings of the 25th Annual International Conference on Computer Science and Software Engineering, pp. 61–69. IBM Corp. (2015)Google Scholar
- 7.Reguly, I.Z., László, E., Mudalige, G.R., Giles, M.B.: Vectorizing unstructured mesh computations for many-core architectures. In: Proceedings of the Programming Models and Applications on Multicores and Manycores, PMAM 2014, pp. 39:39–39:50. ACM, New York (2007)Google Scholar
- 8.Reguly, I.Z., Mudalige, G.R., Bertolli, C., Giles, M.B., Betts, A., Kelly, P.H.J., Radford, D.: Acceleration of a full-scale industrial CFD application with OP 2. IEEE Trans. Parallel Distrib. Syst. PP(99), 1 (2015)Google Scholar
- 9.Reguly, I.Z., Mudalige, G.R., Giles, M.B., Curran, D., McIntosh-Smith, S.: The OPS domain specific abstraction for multi-block structured grid computations. In: Proceedings of the Fourth International Workshop on Domain-Specific Languages and High-Level Frameworks for High Performance Computing, pp. 58–67. IEEE Press (2014)Google Scholar