Segregating heap objects by reference behavior and lifetime

摘要: Dynamic storage allocation has become increasingly important in many applications, part due to the use of object-oriented paradigm. At same time, processor speeds are increasing faster than memory and programs size memories. In this paper, we investigate efforts predict heap object reference lifetime behavior at time objects allocated. Our approach uses profile-based optimization, considers a variety different information sources present object's frequency lifetime. results, based on measurements six intensive programs, show that program references highly predictable our prediction methods can successfully these objects. We decrease page fault rate measured, sometimes dramatically, cases where physical available is constrained.