Category: 

What is a LINPACK Benchmark?

Article Details
  • Written By: Emma G.
  • Edited By: Melissa Wiley
  • Last Modified Date: 13 November 2016
  • Copyright Protected:
    2003-2016
    Conjecture Corporation
  • Print this Article
Free Widgets for your Site/Blog
Snake charmers get snakes to “dance” because of the movement of their flute-like instruments, not their music.  more...

December 4 ,  1945 :  The United States Senate approved of US participation in the United Nations.  more...

A LINPACK benchmark is a way of measuring the speed of a computer. It stands for linear equations software package. There are three different LINPACK benchmarks: LINPACK Fortran n=100, LINPACK n=1000, and the Highly Parallel Computing Benchmark. They compare computer performance when solving a system of linear equations.

There are many different benchmarks used to grade or measure computer speed and performance. The LINPACK benchmark was created in 1979 by Jack Dongarra, a computer science professor at the University of Tennessee. The LINPACK benchmark was created because Dongarra and his associates were working on a book and needed a way to compare computer performance for its readers, so they wrote a program that would run certain mathematical calculations and create a baseline for comparison.

The LINPACK benchmark program was originally written in the Fortran programming language. Fortran means formula translation. It used to translate mathematical formulas into computer code, making it ideal for the LINPACK benchmark application.

There are three different LINPACK benchmarks. LINPACK Fortran n=100 is used to measure a computer's speed at solving linear equations using a matrix or array of numbers containing 100 numbers. In order to return an accurate result, this benchmark must be run exactly as it was written. No changes are allowed even to the comments, which do not directly affect the running of the program.

Ad

The LINPACK n=1000 benchmark is similar to the n=100 benchmark. It is used for a matrix of 1000 numbers. This benchmark must also be run exactly as written.

The Highly Parallel Computing Benchmark is different from the other two benchmarks. Users can specify the problem size and software to be used. This measures the best possible performance of the computer given ideal conditions.

Results of the LINPACK benchmark tests are reported in several ways. They are published in the LINPACK benchmark report, which is subtitled Performance of Various Computers Using Standard Linear Equations Software. It lists the performance of many industry computer systems and can be found online. There is also a website called TOP500 that lists the systems that scored in the top 500 on the benchmarks. The performance of any single computer may vary from the published benchmark results for that model, because every computer is different.

In some applications, LINPACK has been replaced by LAPACK. LAPACK stands for linear algebra package. It restructures the original software for greater efficiency. The program was originally written in Fortran 77 but has now been translated for use in several other programming languages.

Ad

You might also Like

Recommended

Discuss this Article

Post your comments

Post Anonymously

Login

username
password
forgot password?

Register

username
password
confirm
email