Decreasing Memory Overhead in Hard Real-Time Garbage Collection

Abstract

Automatic memory management techniques eliminate many programming errors that are both hard to find and to correct. However, these techniques are not yet used in embedded systems with hard realtime applications. The reason is that current methods for automatic memory management have a number of drawbacks. The two major ones are: (1) not being able to always guarantee short real-time deadlines and (2) using large amounts of extra memory. Memory is usually a scarce resource in embedded applications.In this paper we present a new technique, Real-Time Reference Counting (RTRC) that overcomes the current problems and makes automatic memory management attractive also for hard real-time applications. The main contribution of RTRC is that often all memory can be used to store live objects. This should be compared to a memory overhead of about 500% for garbage collectors based on copying techniques and about 50% for garbage collectors based on mark-and-sweep techniques.KeywordsExecution TimeGarbage CollectionGarbage CollectorMemory OverheadFree ListThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.