Benchmark of the extended Basic Linear Algebra Subprograms on the NEC SX-2 supercomputer
The Basic Linear Algebra Subprograms (BLAS) covering most basic vector operations have been widely used in numerical applications including LINPACK. The Extended Basic Linear Algebra Subprograms (EBLAS) were developed recently as an extension to the strictly vector operations of the BLAS. The EBLAS, also known as Level-2 BLAS, are a set of subprograms that represent basic matrix-vector operations and were written to take advantage of the architecture of today's supercomputers.
The sixteen double precision EBLAS have been installed and benchmarked on the NEC SX-2 Supercomputer at the Houston Area Research Center, The Woodlands, Texas. The EBLAS code, as received, is highly vectorized. The timings and performance rates are presented. Techniques to optimize the EBLAS code on the NEC SX-2 in particular are presented. Performance increases of 10–30% are typical, but dramatic performance increases are obtained for a few routines.
KeywordsSymmetric Matrix Argonne National Laboratory Vector Length Element Vector Original Code
Unable to display preview. Download preview PDF.
- 1.C. L. Lawson, R. J. Hanson, D. R. Kincaid, F. T. Krogh, "Basic Linear Algebra Subprograms for FORTRAN Usage", ACM Transactions on Mathematical Software, (September 1979).Google Scholar
- 2.J. Dongarra, J. Bunch, C. Moler, G. Stewart, "LINPACK Users' Guide", SIAM Publications, Philadelphia, 1979.Google Scholar
- 3.J. Dongarra, S. Hammarling, J. Du Croz, R. Hanson, "An Extended Set of FORTRAN Basic Linear Algebra Subprograms", Argonne Natl. Lab., Technical Memorandum No. 41 Rev. 2, January 1986. (in press ACM Trans. on Math. Software, 1987).Google Scholar
- 4.J.J. Dongarra. "Performance of Various Computers Using Standard Linear Equations Software in a FORTRAN Environment", Argonne National Laboratory, Technical Memorandum No. 43, October, 1986.Google Scholar
- 5.K.G. Stevens, Jr., "Today's Supercomputers", Spring Compcon, pp. 2–6, March 1986.Google Scholar
- 6.O. Lubeck, J. Moore, and R. Mendez. "Benchmark Comparison of the Japanese Supercomputers", International Conference on Supercomputing, Dec. 1985. (update to be published, IEEE Computing, 1987).Google Scholar
- 7.D. H. Bailey, "NASA Kernel Benchmark Results", International Conference of Supercomputing, December 1985. and NEC SX-2 performance data.Google Scholar
- 8.R.W. Hockney. "Parameterization of Computer Performance", invited paper at an International Conference on Vector and Parallel Computing, June 1986, Loen, Norway.Google Scholar
- 9.R. Mendez, "The Performance of the NEC SX-2 Supercomputer System Compared with that of the Cray X-MP/4 and Fujitsu VP-200", Technical Report for Naval Postgraduate School (period April-August 1986), September 1986. (in press Parallel Computing 1987).Google Scholar
- 10.Private communication J. Du Croz, December 1986.Google Scholar
- 11.R.W. Hockney and C.R. Jesshope, "Parallel Computers", Adam Hilger Ltd. Bristol, 1986.Google Scholar
- 12.T. Watanabe, "Architecture and Performance of NEC Supercomputer SX System", International Conference on Vector and Parallel Computing, June 1986, Leon, Norway.Google Scholar