Performance analysis of MPI programs
暂无分享,去创建一个
Possibility to see differences between parallel and sequential parts of the program (that you run on the multiprocessor computer) helps you to predict time that you will need to run this program on the one-processor computer. This time is called useful time. Next, we can calculate the main characteristic of parallel execution – efficiency coefficient that is equal to the ratio of useful time and total time of processor’s using (product of maximum time on a single processor and number of processors). The difference between the whole time of processors using and useful time is lost time. If the programmer doesn’t like his program’s efficiency coefficient he needs to analyze the parts of lost time and the causes of appearance.
[1] Ewing Lusk,et al. Studying parallel program behavior with upshot , 1991 .
[2] William Gropp,et al. Chameleon parallel programming tools users manual , 1993 .
[3] Message P Forum,et al. MPI: A Message-Passing Interface Standard , 1994 .