Identification and removal of program slice criteria for code size reduction in embedded systems

Abstract

Code shrinking is a technique for reducing the size of a software system by detecting and eliminating unused features. It is a static approach that is limited to information known at build time. We demonstrate how to extend code shrinking to take dynamic behavior into account. Our technique, Slice Criteria Identification and REmoval (SCIRE), combines conditional compilation and code shrinking techniques with information provided by the developer about run-time behavior. This integration allows SCIRE to achieve code reduction that is close to optimal while remaining mostly automatic. Our results show that SCIRE reduces code size substantially: In many cases, removal of unused features shrinks middleware footprint to less than 5% of its original size, allowing CORBA and other large software libraries to be used in embedded systems. © International Federation for Information Processing 2007.