To illustrate realistic uses of data parallelism, this example presents two forms of the classic Gauss elimination algorithm for solving systems of linear equations. This particular example is chosen because of the near-universal familiarity with Gaussian elimination, so that maximum attention can be paid to the data parallel techniques with a minimum of distraction from becoming familiar with the problem. One form of the example, called Simple_Gauss, marches the pivot down the main diagonal of the matrix (called Grid or G in the code below); the other form, called Pivot_Gauss, implements the more complicated but more robust maximum pivot strategy for Gaussian elimination.

Both Simple_Gauss and Pivot_Gauss have two versions - a scalar sequential version and a data-parallel version. As presented, the data-parallel version would compile and run; the sequential version is commented out with ``!!" at the beginning of these lines. If these comment characters are removed, and the lines ending with ``!!!!" comments are commented out, the sequential version would compile and run.