The document discusses performance measurement of computers. Benchmark programs are used to standardize performance comparisons between computers. A non-profit organization called the System Performance Evaluation Corporation (SPEC) selects and publishes benchmark application programs from different domains. SPEC ratings are calculated by measuring the time it takes a computer to run each benchmark program, relative to a reference computer system. The overall SPEC rating is the geometric mean of the individual program ratings and provides a measure of a computer's overall performance.
The document discusses performance measurement of computers. Benchmark programs are used to standardize performance comparisons between computers. A non-profit organization called the System Performance Evaluation Corporation (SPEC) selects and publishes benchmark application programs from different domains. SPEC ratings are calculated by measuring the time it takes a computer to run each benchmark program, relative to a reference computer system. The overall SPEC rating is the geometric mean of the individual program ratings and provides a measure of a computer's overall performance.
The document discusses performance measurement of computers. Benchmark programs are used to standardize performance comparisons between computers. A non-profit organization called the System Performance Evaluation Corporation (SPEC) selects and publishes benchmark application programs from different domains. SPEC ratings are calculated by measuring the time it takes a computer to run each benchmark program, relative to a reference computer system. The overall SPEC rating is the geometric mean of the individual program ratings and provides a measure of a computer's overall performance.
The document discusses performance measurement of computers. Benchmark programs are used to standardize performance comparisons between computers. A non-profit organization called the System Performance Evaluation Corporation (SPEC) selects and publishes benchmark application programs from different domains. SPEC ratings are calculated by measuring the time it takes a computer to run each benchmark program, relative to a reference computer system. The overall SPEC rating is the geometric mean of the individual program ratings and provides a measure of a computer's overall performance.
Download as DOCX, PDF, TXT or read online from Scribd
Download as docx, pdf, or txt
You are on page 1/ 2
PERFORMANCE MEASUREMENT
It is important to be able to assess the performance of a computer.
Computer designers use performance estimates to evaluate the effectiveness of new features. Manufacturers use performance indicators in the marketing process. Buyers use such data to choose among many available computer models. The previous discussion suggests that the only parameter that properly describes the performance of a computer is the execution time, T, for the programs of interest. Despite the conceptual simplicity of Equation, computing the value of T is not simple. Moreover, parameters such as the clock speed and various architectural features are not reliable indicators of the expected performance. For these reasons, the computer community adopted the idea of measuring computer performance using benchmark programs. To make comparisons possible, standardized programs must be used. The performance measure is the time it takes a computer to execute a given benchmark. Initially, some attempts were made to create artificial programs that could be used as standard benchmarks. But, synthetic programs do not properly predict performance obtained when real application programs are run. The accepted practice today is to use an agreed upon selection of real application programs to evaluate performance. The accepted practice today is to use an agreed-upon selection of real application programs to evaluate performance. A nonprofit organization called System Performance Evaluation Corporation (SPBC) selects and publishes representative application programs for different application domains, together with test results for many commercially available computers. For general-purpose computers, a suite of benchmark programs was selected in 1989. It was modified somewhat and published in 1995 and again in 2000.
The programs selected range from game playing, compiler, and database applications to numerically intensive programs in astrophysics and quantum chemistry. In each case, the program is compiled for the computer under test, and the running time on a real computer is measured. The same program is also compiled and run on one computer selected as a reference. For SPEC95, the reference is the SUN SPARCstation 10/40. For SPEC2000, the reference computer is an Ultra- SPARCIO workstation with a 300-MHz UltraSPARC-Ili processor. The SPEC rating is computed as follows: Thus a SPEC rating of 50 means that the computer under test is 50 times as fast as the UltraSPARCIO for this particular benchmark. The test is repeated for all the programs in the SPEC suite, and the geometric mean of the results is computed. Let SPECi, be the rating for program in the suite. The overall SPEC rating for the computer is given by: where n is the number of programs in the suite. Because the actual execution time is measured, the SPEC rating is a measure of the combined effect of all factors affecting performance, including the compiler, the operating system, the processor, and the memory of the computer being tested. Details about the SPEC benchmark programs and results of the tests conducted can be found on the SPEC web page.
