The key to the creation of more efficient software for mobile platforms and multicore chips could lie in artificial intelligence (AI), and the MilePost project seeks to make this vision a reality. The project has devised an experimental version of the GNU Compiler Collection that employs AI to enhance the quality of its own output so that compiler developers can spend less time modifying compilers for particular platforms by allowing the compilers to do that by themselves.
MilePost utilizes machine-learning methods to collect data on software performance and make appropriate adjustments to its outputted machine code. The compiler examines the source code input to find specific "features" that might be suitable candidates for optimization. Once a catalog of all the features present in a given program is organized, MilePost can use statistical methods to decide which optimizations will generate the best results and tweak its own modular design as appropriate.
Early tests by IBM demonstrate that MilePost can upgrade performance by up to 18 percent compared to traditional compilers' code output. Code optimization becomes vital when focusing on mobile devices and other gadgets with low-powered processors and limited resources. Besides benefiting mobile devices, self-modifying compilers could help optimize software for multicore processors.
View Full Article
Abstracts Copyright © 2009 Information Inc., Bethesda, Maryland, USA
No entries found