Communications of the ACM
When Good Enough is Better
Credit: Christine Daniloff / MIT News
Massachusetts Institute of Technology (MIT) researchers have developed a system that automatically finds parts of computer code where accuracy can be traded for significant increases in speed. In one set of tests, the system analyzed video data for transmission over the Internet and cut the coding time in half with no noticeable effect on the video quality. The researchers say the approach could have advantages for any system that needs to process data in real time.
The system also could benefit programs that need to look for patterns in huge masses of data, such as website recommendation engines.
The MIT system minimizes a common computer programming process called a loop. The researchers call the new technique "loop perforation" because it punches holes in loops by skipping every other step in the loop, or as many steps as possible without sacrificing too much accuracy. The system searches through a program, perforating each loop in turn, executing the program, and using standard measures to gauge the effect on performance. It then determines which loops' perforation provides the greatest increases in speed with the smallest drop in performance.
From MIT News
View Full Article
Abstracts Copyright © 2010 Information Inc., Bethesda, Maryland, USA 
No entries found