Abstract
The use of highly optimized inner kernels is of paramount importance for obtaining efficient numerical algorithms. Often, such kernels are created by hand. In this paper, however, we present an alternative way to produce efficient matrix multiplication kernels based on a set of simple codes which can be parameterized at compilation time. Using the resulting kernels we have been able to produce high performance sparse and dense linear algebra codes on a variety of platforms.
This work was supported by the Ministerio de Ciencia y Tecnología of Spain (TIN2004-07739-C02-01).
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
Kamath, C., Ho, R., Manley, D.: DXML: A high-performance scientific subroutine library. Digital Technical Journal 6, 44–56 (1994)
Navarro, J.J., García, E., Herrero, J.R.: Data prefetching and multilevel blocking for linear algebra operations. In: Proceedings of the 10th international conference on Supercomputing, pp. 109–116. ACM Press, New York (1996)
Anderson, E., Bai, Z., Dongarra, J., Greenbaum, A., McKenney, A., Croz, J.D., Hammarling, S., Demmel, J., Bischof, C., Sorensen, D.: LAPACK: A portable linear algebra library for high-performance computers. In: Proc. of Supercomputing 1990, pp. 1–10. IEEE Press, Los Alamitos (1990)
Kåagström, B., Ling, P., van Loan, C.: Gemm-based level 3 blas: high-performance model implementations and performance evaluation benchmark. ACM Transactions on Mathematical Software (TOMS) 24, 268–302 (1998)
Dongarra, J.J., Du Croz, J., Duff, I.S., Hammarling, S.: A set of level 3 basic linear algebra subprograms. ACM Trans. Math. Software 16, 1–17 (1990)
Bilmes, J., Asanovic, K., Chin, C.W., Demmel, J.: Optimizing matrix multiply using PHiPAC: a portable, high-performance, ANSI C coding methodology. In: 11th ACM Int. Conf. on Supercomputing, pp. 340–347. ACM Press, New York (1997)
Whaley, R.C., Dongarra, J.J.: Automatically tuned linear algebra software. In: Supercomputing 1998, pp. 211–217. IEEE Computer Society, Los Alamitos (1998)
Bacon, D.F., Graham, S.L., Sharp, O.J.: Compiler transformations for highperformance computing. ACM Computing Surveys 26, 345–420 (1994)
Herrero, J.R., Navarro, J.J.: Automatic benchmarking and optimization of codes: an experience with numerical kernels. In: Int. Conf. on Software Engineering Research and Practice, pp. 701–706. CSREA Press (2003)
Intel: Intel(R) Itanium(R) 2 processor reference manual for software development and optimization (2004)
Fuchs, G., Roy, J., Schrem, E.: Hypermatrix solution of large sets of symmetric positive-definite linear equations. Comp. Meth. Appl. Mech. Eng. 1, 197–216 (1972)
Noor, A., Voigt, S.: Hypermatrix scheme for the STAR–100 computer. Comp. & Struct. 5, 287–296 (1975)
Herrero, J.R., Navarro, J.J.: Improving performance of hypermatrix cholesky factorization. In: Kosch, H., Böszörményi, L., Hellwagner, H. (eds.) Euro-Par 2003. LNCS, vol. 2790, pp. 461–469. Springer, Heidelberg (2003)
Lam, M., Rothberg, E., Wolf, M.: The cache performance and optimizations of blocked algorithms. In: Proceedings of ASPLOS 1991, pp. 67–74 (1991)
Herrero, J.R., Navarro, J.J.: Adapting linear algebra codes to the memory hierarchy using a hypermatrix scheme. In: Int. Conf. on Parallel Processing and Applied Mathematics (2005)
Daydé, M.J., Duff, I.S.: The use of computational kernels in full and sparse linear solvers, efficient code design on high-performance RISC processors. In: VECPAR, pp. 108–139 (1996)
SSE2 (Streaming SIMD Extensions 2 for the Pentium 4 processor), http://www.intel.com/software/products/college/ia32/sse2
Gunnels, J.A., Henry, G., van de Geijn, R.A.: A family of high-performance matrix multiplication algorithms. In: International Conference on Computational Science (1), pp. 51–60 (2001)
Navarro, J.J., Juan, A., Lang, T.: MOB forms: A class of Multilevel Block Algorithms for dense linear algebra operations. In: Proceedings of the 8th International Conference on Supercomputing. ACM Press, New York (1994)
Goto, K., van de Geijn, R.: On reducing TLB misses in matrix multiplication. Technical Report CS-TR-02-55, Univ. of Texas at Austin (2002)
Gustavson, F.G.: New generalized data structures for matrices lead to a variety of high performance algorithms. In: Wyrzykowski, R., Dongarra, J., Paprzycki, M., Waśniewski, J. (eds.) PPAM 2001. LNCS, vol. 2328, pp. 418–436. Springer, Heidelberg (2002)
Andersen, B.S., Wasniewski, J., Gustavson, F.G.: A recursive formulation of Cholesky factorization of a matrix in packed storage. ACM Transactions on Mathematical Software (TOMS) 27, 214–244 (2001)
Chatterjee, S., Lebeck, A.R., Patnala, P.K., Thottethodi, M.: Recursive array layouts and fast parallel matrix multiplication. In: Proc. of the 11th annual ACM symposium on Parallel algorithms and architectures, pp. 222–231. ACM Press, New York (1999)
Valsalam, V., Skjellum, A.: A framework for high-performance matrix multiplication based on hierarchical abstractions, algorithms and optimized low-level kernels. Concurrency and Computation: Practice and Experience 14, 805–839 (2002)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2006 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Herrero, J.R., Navarro, J.J. (2006). Compiler-Optimized Kernels: An Efficient Alternative to Hand-Coded Inner Kernels. In: Gavrilova, M.L., et al. Computational Science and Its Applications - ICCSA 2006. ICCSA 2006. Lecture Notes in Computer Science, vol 3984. Springer, Berlin, Heidelberg. https://doi.org/10.1007/11751649_84
Download citation
DOI: https://doi.org/10.1007/11751649_84
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-34079-9
Online ISBN: 978-3-540-34080-5
eBook Packages: Computer ScienceComputer Science (R0)