Results of Profiling and Analysis of Behaviour of Memory Objects in Java

The performance of a memory system depends to great extent on the algorithms used for memory allocation and garbage collection (GC). GC implementers need accurate data about memory objects' behaviour in order to produce high quality garbage collectors optimized for their particular problems. Therefore, different GC-related data should be collected and analysed afterwards. We propose a set of parameters that should be traced in benchmark programs and organization of an analysis environment. Then we present several examples of results obtainable by the proposed analysis environment. These results can be useful in improving the existing or in the creation of new GC algorithms . The profiled data are obtained by running benchmarks written in the Java programming language that we use as a representative of the languages that belong to the object oriented programming paradigm.