PaCT 2003: Parallel Computing Technologies pp 345-353 | Cite as
Analysis of Architecture and Design of Linear Algebra Kernels for Superscalar Processors
Abstract
In this paper we present methods for developing high performance computational kernels and dense linear algebra routines. First, the microarchitecture of AMD Athlon processors is analyzed, with the goal to achieve peak computational rates. These processors are widely used for building inexpensive PC clusters. Then, different approaches for implementing matrix multiplication algorithms are analyzed for hierarchical memory computers, taking into account their architectural properties and limitations. Block versions of matrix multiplication and LU-decomposition algorithms are considered. Finally, the obtained performance results for AMD Athlon/Duron processors are discussed in comparison with other approaches.
Keywords
instruction level parallelism microarchitecture out-of-order processors cache memories linear algebra kernels performance measurements LINPACK benchmarkPreview
Unable to display preview. Download preview PDF.
References
- 1.Dongarra, J., Walker, D.: The Design of Linear Algebra Libraries for High Performance Computers. LAPACK Working Note 58. University of Tennessee, Knoxville, TN (1993)Google Scholar
- 2.Bessonov, O., Fougère, D., Dang Quoc, K., Roux, B.: Methods for Achieving Peak Computational Rates for Linear Algebra Operations on Superscalar RISC Processors. In: Malyshkin, V.E. (ed.) PaCT 1999. LNCS, vol. 1662, pp. 180–185. Springer, Heidelberg (1999)CrossRefGoogle Scholar
- 3.Dongarra, J.: Performance of Various Computers Using Standard Linear Equations Software. Report CS-89-85. University of Tennessee, Knoxville, and ORNL, Oak Ridge, TN (2003)Google Scholar
- 4.Whaley, R.C., Petitet, A., Dongarra, J.: Automated Empirical Optimization of Software and the ATLAS Project. Parallel Computing 27(1-2), 3–35 (2001)CrossRefMATHGoogle Scholar
- 5.AMD Athlon TM Processor x86 Code Optimization Guide. Advanced Micro Devices, Publication No. 22007 (February 2002) Google Scholar
- 6.Ortega, J.M.: Introduction to Parallel and Vector Solution of Linear Systems. Plenum Press, New York (1988)Google Scholar
- 7.Chen, Z., Dongarra, J., Luszczek, P., Roche, K.: Self Adapting Software for Numerical Linear Algebra and LAPACK for Clusters. LAPACK Working Note 160. University of Tennessee, Knoxville, TN (2003)Google Scholar
- 8.Software Optimization Guide for AMD Athlon TM 64 and AMD Opteron TM Processors. Advanced Micro Devices, Publication No. 25112 (April 2003)Google Scholar