JAVA VERSION OF LINPACK BENCHMARK OFFERED AT THE UNIVERSITY OF TENNESSEE

Standardized, rigorous, and scientifically tenable methodology for studying the performance of computer systems continues to be of great value to the high-performance computing community. The Linpack Benchmark is one such numerically intensive test that has been used for years to measure the floating point performance of computers. Through their Netlib Web site, University of Tennessee CRPC researchers Jack Dongarra and Reed Wade now offer an applet that runs the Linpack Benchmark in Java.

Users of the applet can see the performance of their computer and compare it with other machines running the Java applet. The benchmark solves a dense 100 x 100 system of linear equations with one right-hand side, Ax = b. The matrix is generated randomly and the right-hand side is constructed so the solution has all components equal to one. The method of solution is based on Gaussian elimination with partial pivoting. Once the problem is solved on a user's computer, a comparison is given with other computers running the same benchmark. The timing information can then be sent back to Tennessee and entered into the timing collection.

To run the Linpack Benchmark in Java, see http://www.netlib.org/benchmark/linpackjava /. The Java code that implements the benchmark can be viewed as well.

For more information, contact Jack Dongarra at dongarra@cs.utk.edu .

Table of Contents