A Flexible, Efficient Concurrent Garbage Collector for Speculative Thread Processors

Michael Chen and Kunle Olukotun Computer Systems Lab, Stanford University Abstract In this paper, we introduce a novel garbage collector for Java to be used for processors with speculative threads support like the Hydra chip multiprocessor (CMP). Thread speculation permits parallel execution of sections of sequential code with data dependencies enforced in the hardware, eliminating the need for explicit locking. We have augmented Dijkstra’s classical on-the-fly mark and sweep collector to take advantage of the CMP’s thread-level speculation and low-latency interprocessor communication. The resulting collector can execute concurrently without an explicit collector thread or can increase the collection rate by executing in parallel on all free processors. A dynamically scalable collector provides more control to adjust collection behavior according to the total system load and real-time deadlines. Speculative threads ensured the scalable implementation was simple and required no additional overheads, but we were unable to observe additional performance gains from eliminated explicit synchronization.