Mission
Many of the most intriguing problems that modern science attempts to solve cannot be easily verified by direct experimentation. These problems include the study of the earliest moments in the history of our universe, gaining an understanding of controlled fusion reactions, or predicting changes in the earth's climate over the next 100 years. Each problem's solution can be best approached using powerful computers programmed to simulate environments and physical scales not amenable to laboratory experimentation.
Simulation runs for such problems require the prodigious speed, memory and data storage resources available only with the world's most capable computers. The number and diversity of high-performance computers, as well as the vast breadth of applications run on them, make it difficult to compare one machine with another.
The Applications Performance Matrix project provides a rich source of information on the performance of high-performance computers applied to real science and engineering problems. The scientist can learn what type of computational approach is currently the most capable for solving the problem of interest; the computer software expert can compare the performance of a computational algorithm against a competing algorithm; and the hardware specialist can determine how processor speed, architecture, or number of processors affects computational performance in real applications.
We see the Applications Performance Matrix as a community effort. We actively solicit input from the users of high-performance machines. We also seek your advice and opinions about the sort of information the Applications Performance Matrix should contain and the form in which it is best presented. Please join us in this effort.
