nokia and csail

Home    About NRCC    Research Projects    Publications    Image Gallery    Blog  

Overview: ARMO

There is an increasing need to support a variety of applications on mobile phones, and many will be developed by Independent Software Vendors. This need is transforming the perception of a mobile phone from an embedded device into a stable platform for software development. Nevertheless, Mobile phones will continue to have an "embedded device" character because the device must be small, dissipate less than 3 watts maximum, and be able to operate for days on battery power. This dual characteristic of a mobile phone as a platform and as an embedded device provides many research challenges. Because of the stringent performance and power requirements, a mobile system might be implemented with a mixture of hardware and software components, and software components could be partitioned across multiple parallel general-purpose processors or DSPs. The appropriate implementation of a component could even change from hardware to software depending on performance and power requirements. Perhaps the right way to think about designing a mobile platform will not be as a general-purpose processor programmed using serial languages and libraries of APIs, but rather as a collection of concurrently executing modules which support the right level of functionality for a given market.

ARMO stands for ARchitecture MOdeling (in Finnish it means Grace!). The principal underlying technology in this project is a modular system implementation methodology based on Guarded Atomic Actions and Decoupled Systems. This methodology allows a single module specification to be automatically translated into high-quality hardware implementations or high-quality software implementations, with automatic generation of interfaces between the domains. This vision will enable a complete mobile phone to be "assembled" rapidly from existing or new modules, where each module is mapped to either hardware or software according to performance and power requirements. We will show via example implementations that a design methodology based on Guarded Atomic Actions and Decoupled Systems allows packaging of Intellectual Property in a way that is highly reusable because of parameterization and flexible interfaces. We expect our design flow to dramatically reduce the design cost and risk for implementing custom hardware blocks that interact seamlessly with software modules. An important component of this project is to quantify the performance and power consumption of different implementation technologies used to implement computationally intensive functions required in a mobile terminal.

To support all of these efforts, we will build a full-system modeling framework, where different components can be simulated at different levels of abstraction (functional, transactional, cycle-level). This may also result in a non-proprietary model of a mobile phone that can be shared with other academic and outside researchers. Within this proposal, the full system model will be used to explore new capabilities in mobile platform architecture, including hardware security mechanisms and high-bandwidth wireless links.

CSAIL    Nokia Research
Media inquiries email: news@csail.mit.edu