Abstract
Commodity clusters revolutionized high performance computing in a fundamental way since its first inception, and now now dominate many of the world’s premiere supercomputers on the Top500, growing to the scale of over 10,000 CPU cores and beyond. Still,"classical" specialized vector supercomputers still remain to be sold and facilitated, especially at high end of the market, largely due to the nature of some of the HPC workloads still requiring the computing power of vectors, in areas such as CFD, FEM with kernels such as FFT, characterized as mostly large-scale irregular sparse codes Finally, however, commoditization of vector computing is on the rise, lead by multimedia application requirements, and spurred many architectures to arise such as GPUs and the Cell processor. But various problems still remain by which we cannot claim with 100% confidence that commodity vectors are here to stay in the HPC space. Research and development has to be conducted at various degrees of intensity to utilize the new breed of commodity vector hardware to their fullest capabilities, just as various research were needed to harness the power and the scalability of commodity clusters. In the talk I will outline some of the details, and our recent research endeavors aimed at solving the various issues.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsPreview
Unable to display preview. Download preview PDF.
References
Meuer, H.W., Strohmaier, E., Dongarra, J.J., Simon, H.D.: The Top500 Supercomputer Sites (1993), http://www.top500.org
Owens, J., Houston, M., Luebke, D., Green, S., Stone, J., Phillips, J.: GPU Computing. In: Proceesings of the IEEE, vol. 96(5), pp. 879–899. The IEEE Press, Los Alamitos (2008)
Shirts, M., Pande, V.: Screen Savers of the World Unite! The Science Magazine, the American Association for the Advancement of Science 290(2498), 1903–1904 (2000), http://folding.stanford.edu/
Chen, T., Raghavan, R., Dale, J.N., Iwata, E.: Cell Broadband Engine Architecture and its first implementation—A performance view. IBM Journal of Research and Development 51(5) (August 2007)
Clearspeed Whitepaper: CSX Processor Architecture, http://www.clearspeed.com/docs/resources/ClearSpeed_Architecture_Whitepaper_Feb07v2.pdf
Endo, T., Matsuoka, S.: Massive Supercomputing Coping with Heterogeneity of Modern Accelerators. In: Proc. IEEE International Parallel and Distributed Processing Symposium. The IEEE CS Press, Los Alamitos (2008) (CD-ROM)
Susukita, R., Ebisuzaki, T., et al.: Hardware accelerator for molecular dynamics: MDGRAPE-2. Computer Physics Communications 155(2), 115–131 (2003)
To, Y., Furui, T., Nishikawa, T., Yamamoto, M., Inoue, K.: YA Development Concept of Supercomputer SX-8. NEC Technical Journal (In Japanese) 58(4), 3–6 (2005)
Ohshima, S., Kise, K., Katagiri, T., Yuba, T.: Parallel processing of matrix multiplication in a cpu and gpu heterogeneous environment! In: Daydé, M., Palma, J.M.L.M., Coutinho, Á.L.G.A., Pacitti, E., Lopes, J.C. (eds.) VECPAR 2006. LNCS, vol. 4395, pp. 305–318. Springer, Heidelberg (2007)
Fujimoto, N.: Faster Matrix-Vector Multiplication on GeForce 8800GTX. In: Proc. LSPP Workshop 2008, IPDPS 2008 Proceedings. The IEEE CS Press, Los Alamitos (2008) (CD-ROM)
Ogata, Y., Endo, T., Maruyama, N., Matsuoka, S.: An Efficient, Model-Based CPU-GPU Heterogeneous FFT Library. In: Proc. HCW 2008: 17th International Heterogeneity in Computing Workshop, IPDPS 2008 Proceedings. The IEEE CS Press, Los Alamitos (2008) (CD-ROM)
Himeno, R., et al.: The Himeno Benchmark Contest, Riken (2002), http://accc.riken.jp/HPC/HimenoBMT/contest_e.html
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2008 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Matsuoka, S. (2008). The Rise of the Commodity Vectors. In: Palma, J.M.L.M., Amestoy, P.R., Daydé, M., Mattoso, M., Lopes, J.C. (eds) High Performance Computing for Computational Science - VECPAR 2008. VECPAR 2008. Lecture Notes in Computer Science, vol 5336. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-92859-1_7
Download citation
DOI: https://doi.org/10.1007/978-3-540-92859-1_7
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-92858-4
Online ISBN: 978-3-540-92859-1
eBook Packages: Computer ScienceComputer Science (R0)