IBM has announced a collaboration with European Union partners to develop new software that will hopefully improve performance and cut down time-to-market of mobile web applications. Specifically, partners of the Milepost project ("MachIne Learning for Embedded PrOgramS opTimization") are developing advanced artificial intelligence technology that automatically learns how to best optimize newly developed programs for embedded processors in mobile devices. Milepost project members include IBM Haifa Research Lab, Israel; the University of Edinburgh, UK; ARC International Ltd., UK; CAPS Enterprise, France; and INRIA, France
In a recently released study, IBM claims that 80 percent of consumers would prefer a service provider that gave them more choice in the applications and services available on their mobile device. For developers, the study points to how consumer demand for customization and personalization will drive the need for projects like Milepost to enable a faster path to market for new mobile applications.
Today's mobile hardware designs are rapidly changing and current hand-crafted approaches to mobile software development are no longer sustainable. The project's partners have released a prototype version of their software at the GCC Summit showcasing successful preliminary results. Within one month, Milepost was able to improve the performance of a state-of-the-art complier by 10 percent -- something that would normally take several years to accomplish.
"The Milepost solution uses artificial intelligence and machine learning to understand what kind of compiler optimizations are optimal for use with each new hardware design," explained Mike O'Boyle, Professor of Computer Science at the University of Edinburgh and Project Coordinator for Milepost. "This will help completely automate compiler construction and enable more rapid code design of hardware and software -- dramatically reducing time to market in these systems."
With each generation of reconfigurable architecture, the compiler development time increases and the performance improvement achieved is at risk. As high performance embedded systems move from application specific ASICs to programmable heterogeneous processors, this problem is becoming critical. Compiler designs simply can't keep up with so many different kinds of new processors.
"Milepost is realizing the vision of customized hardware with tailor fit software," noted Dr. Bilha Mendelson, Manager of Code Optimization Technologies at the IBM Haifa Research Lab. "Aside from shorter design cycles, Milepost opens new opportunities by enabling engineers to leap ahead and work with more experimental hardware. Opening the compiler infrastructure and combining it with machine learning techniques enables us to generate several sets of optimizations sequences for each hardware application area."
Milepost partners are in the midst of a three-year program, at the end of which, they expect to release a fully robust version of their compiler optimization software into the GCC main product. The Milepost GCC version will be available to everyone in the open source community and is scheduled for release in June 2009.