Abstract：Most existing garbage collection methods unilaterally emphasize the performance in space or time,and cannot adjust themselves dynamically according to the actual situation.This paper that combines the garbage collection methods of Ungar and Shaw proposes a new self-adaptive generational garbage collection method.A valve line counted from residual object statistics table is used to reduce cost of collecting estimate,and big object regions is proposed to reduce the big objects' movements,furthermore,smaller middle generations is presented to avoid upgrading earlier.Experiment results show that this method does improve the efficiency of garbage collection.
