Tools for Performance Measurement

The reason behind these algorithms has been to gain a speed up and improve the performance. After the parallel algorithm has been written and implemented, the performance of both the algorithms is one of the other main concerns. In order to analyze the performance of algorithms, there are many kinds of performance measurement tools. The measurement tools depend not only on the parallel algorithm but need to collect data from the operating system and the hardware being used, so as to give effective utilization of the tools.

For a given parallel computer, as the computational power and its number of processors increases, the volume and complexity of data for performance analysis substantially increases. The gathered data for measurement is always very hard for the tools to store and process it.

The task of calculating the performance of the parallel programs has been divided into two components.

1) Performance Analysis:  It gives the vital information to the programmers from the large chunk of statistics available of the program while in implementation mode or from the output data.

2) Performance Instrumentation:  Its emphasis on how to efficiently gather information about the computation of the parallel computer.