The Active Memory Module: Bounded Garbage Collection Latency for Real-time Embedded Systems

In this paper, the performance analysis of the proposed Active Memory Module (AMM) for embedded systems is presented. Unlike the software counterparts, the AMM can perform a memory allocation in a predictable and bounded fashion (5 cycles). Moreover, it can also yield a bounded sweeping time regardless of the number of live objects or heap size. By utilizing the proposed system, the overall speedup can be as high as 23% comparing to the garbage collection system of the JDK 1.2.2 running in classic mode. Additionally, an incremental marking approach is also presented. The proposed scheme is a hybrid between reference counting and mark-sweep. One-bit reference-count field is used to assist incremental marking. A full mark-sweep is used to collect cyclic structures and dead objects with stuck reference bit. The proposed scheme takes full advantage of constant time sweeping provided by the hardware. At the same time, it can also predict the tracing latency prior to a full collection cycle. By using the proposed scheme, the time to reclaim dead objects can be accurately predicted.