Abstract
The full format data structures of Dense Linear Algebra hurt the performance of its factorization algorithms. Full format rectangular matrices are the input and output of level the 3 BLAS. It follows that the LAPACK and Level 3 BLAS approach has a basic performance flaw. We describe a new result that shows that representing a matrix A as a collection of square blocks will reduce the amount of data reformating required by dense linear algebra factorization algorithms from O(n 3) to O(n 2). On an IBM Power3 processor our implementation of Cholesky factorization achieves 92% of peak performance whereas conventional full format LAPACK DPOTRF achieves 77% of peak performance. All programming for our new data structures may be accomplished in standard Fortran, through the use of higher dimensional full format arrays. Thus, new compiler support may not be necessary. We also discuss the role of concatenating submatrices to facilitate hardware streaming. Finally, we discuss a new concept which we call the L1 / L0 cache interface.
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
Preview
Unable to display preview. Download preview PDF.
References
Agarwal, R.C., Gustavson, F.G., Zubair, M.: Exploiting functional parallelism of POWER2 to design high-performance numerical algorithms. IBM Journal of Research and Development 38(5), 563–576 (1994)
Andersen, B.S., Gustavson, F.G., Wasnieski, J.: A Recursive Formulation of Cholesky Factorization of a Matrix in Packed Storage. ACM TOMS 27(2), 214–244 (2001)
Andersen, B.S., Gunnels, J.A., Gustavson, F.G., Reid, J.K., Wasnieski, J.: A Fully Portable High Performance Minimal Storage Hybrid Cholesky Algorithm. ACM TOMS 31(2), 201–227 (2005)
Anderson, E., Bai, Z., Bischof, C., Demmel, J., Dongarra, J., Du Croz, J., Greenbaum, A., Hammarling, S., McKenney, A., Ostrouchov, S., Sorensen, D.: LAPACK Users’ Guide Release 3.0, SIAM, Philadelphia (1999), http://www.netlib.org/lapack/lug/lapack_lug.html
Bilmes, J., Asanovic, K., Whye Chin, C., Demmel, J.: Optimizing Matrix Multiply Using PHiPAC: A Portable, High-Performance, ANSI C Coding Methodology. In: Proceedings of International Conference on Supercomputing, Vienna, Austria (1997)
Chatterjee, S., et al.: Design and Exploitation of a High-performance SIMD Floating-point Unit for Blue Gene/L. IBM Journal of Research and Development 49(2-3), 377–391 (2005)
Dongarra, J.J., Moler, C.B., Bunch, J.R., Stewart, G.W.: LINPACK Users’ Guide Release 2.0. SIAM, Philadelphia (1979)
Dongarra, J.J., Gustavson, F.G., Karp, A.: Implementing Linear Algebra Algorithms for Dense Matrices on a Vector Pipeline Machine. SIAM Review 26(1), 91–112 (1984)
Dongarra, J.J., Du Croz, J., Hammarling, S., Hanson, R.J.: An Extended Set of FORTRAN Basic Linear Algebra Subprograms. TOMS 14(1), 1–17 (1988)
Dongarra, J.J., Du Croz, J., Hammarling, S., Duff, I.: A Set of Level 3 Basic Linear Algebra Subprograms. TOMS 16(1), 1–17 (1990)
Elmroth, E., Gustavson, F.G., Kagstrom, B., Jonsson, I.: Recursive Blocked Algorithms and Hybrid Data Structures for Dense Matrix Library Software. SIAM Review 46(1), 3–45 (2004)
Gunnels, J., Gustavson, F.G., Henry, G., van de Geijn, R.: Formal linear algebra methods environment (FLAME). ACM TOMS 27(4), 422–455 (2001)
Gunnels, J.A., Gustavson, F.G.: A New Array Format for Symmetric and Triangular Matrices. In: Dongarra, J.J., Madsen, K., Waśniewski, J. (eds.) PARA 2004. LNCS, vol. 3732, pp. 247–255. Springer, Heidelberg (2006)
Gunnels, J.A., Gustavson, F.G., Henry, G.M., van de Geijn, R.A.: A Family of High-Performance Matrix Multiplication Algorithms. In: Dongarra, J.J., Madsen, K., Waśniewski, J. (eds.) PARA 2004. LNCS, vol. 3732, pp. 256–265. Springer, Heidelberg (2006)
Gustavson, F.G.: Recursion Leads to Automatic Variable Blocking for Dense Linear-Algebra Algorithms. IBM Journal of Research and Development 41(6), 737–755 (1997)
Gustavson, F.G., Jonsson, I.: Minimal Storage High Performance Cholesky via Blocking and Recursion. IBM Journal of Research and Development 44(6), 823–849 (2000)
Gustavson, F.G.: High Performance Linear Algebra Algorithms using New Generalized Data Structures for Matrices. IBM Journal of Research and Development 47(1), 31–55 (2003)
Gustavson, F.G.: New Generalized Data Structures for Matrices Lead to a Variety of High performance Dense Linear Algorithms. In: Dongarra, J.J., Madsen, K., Waśniewski, J. (eds.) PARA 2004. LNCS, vol. 3732, pp. 11–20. Springer, Heidelberg (2006)
Gustavson, F.G., Wasniewski, J.: Rectangular Full Packed Format for LAPACK Algorithms Timings on Several Computers. In: Kagstom, B., Elmroth, E. (eds.) Para 2006. LNCS, vol. xxxx, pp. 570–579. Springer, Heidelberg (2006)
IBM: IBM Engineering and Scientific Subroutine Library for AIX Version 3, Release 3. IBM Pub. No. SA22-7272-04 (December 2001)
Kalla, R., Sinharoy, B., Tendler, J.: Power 5. HotChips-15, August 17-19, 2003, Stanford, CA (2003)
Lawson, C.L., Hanson, R.J., Kincaid, D.R., Krogh, F.T.: Basic Linear Algebra Subprograms for Fortran Usage. TOMS 5(3), 308–323 (1979)
Park, N., Hong, B., Prasanna, V.K.: Tiling, Block Data Layout, and Memory Hierarchy Performance. IEEE Trans. Parallel and Distributed Systems 14(7), 640–654 (2003)
Sinharoy, B., Kalla, R.N., Tendler, J.M, Kovacs, R.G., Eickemeyer, R.J., Joyner, J.B.: POWER5 System Microarchitecture. IBM Journal of Research and Development 49(4/5), 505–521 (2005)
Whaley, R.C., Petitet, A., Dongarra, J.J.: Automated Empirical Optimization of Software and the ATLAS Project. Parallel Computing (1-2), 3–35 (2001)
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 2007 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Gustavson, F.G., Gunnels, J.A., Sexton, J.C. (2007). Minimal Data Copy for Dense Linear Algebra Factorization. In: Kågström, B., Elmroth, E., Dongarra, J., Waśniewski, J. (eds) Applied Parallel Computing. State of the Art in Scientific Computing. PARA 2006. Lecture Notes in Computer Science, vol 4699. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-75755-9_66
Download citation
DOI: https://doi.org/10.1007/978-3-540-75755-9_66
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-75754-2
Online ISBN: 978-3-540-75755-9
eBook Packages: Computer ScienceComputer Science (R0)