Java Memory Allocation with Lazy Worst Fit for Small Objects

Hyung-Kyu Choi, Yoo C. Chung, Soo-Mook Moon

Abstract: Memory allocation is an important part of modern programming languages, including garbage-collected languages such as Java. We propose a fast memory allocation scheme for Java using lazy worst fit (LWF), where pointer increment is used as the primary allocation method and worst fit is used as a backup. We evaluated LWF on a working Java virtual machine with non-moving garbage collection, and the results show that LWF is practically useful since the overhead of fit allocation and the amount of fragmentation are low.