Author

Xiu Guo

Publication Date

Fall 2012

Degree Type

Master's Project

Department

Computer Science

Abstract

As predicted by Moore's law, the number of transistors on a chip has been doubled approximately every two years. As miraculous as it sounds, for many years, the extra transistors have massively benefited the whole computer industry, by using the extra transistors to increase CPU clock speed, thus boosting performance. However, due to heat wall and power constraints, the clock speed cannot be increased limitlessly. Hardware vendors now have to take another path other than increasing clock speed, which is to utilize the transistors to increase the number of processor cores on each chip. This hardware structural change presents inevitable challenges to software structure, where single thread targeted software will not benefit from newer chips or may even suffer from lower clock speed. The two fundamental challenges are: 1. How to deal with the stagnation of single core clock speed and cache memory. 2. How to utilize the additional power generated from more cores on a chip. Most software programming languages nowadays have distributed computing support, such as C and Java [1]. Meanwhile, some new programming languages were invented from scratch just to take advantage of the more distributed hardware structures. The X10 Programming Language is one of them. The goal of this project is to evaluate X10 in terms of performance, programmability and tool support.

Share

COinS