Mark-sweep Collector Research Articles

Mark without much Sweep Algorithm for Garbage Collection

In this paper two simple improvements over traditional mark-sweep collector are proposed. The core idea is placing small objects of the same type in buckets. The buckets are organised in such way to eliminate the internal fragmentation, sweeping, and freeing inside them. The measured improvement of garbage collection time over traditional mark-sweep is 19%. Another proposed improvement is more general and is applicable to other garbage collection algorithms as well. It uses heuristics to control the heap growth. The regularities in behaviour of objects of particular types are used to determine whether the collection should be performed or avoided in favour of immediate heap expansion. The heap expansion algorithm reduces garbage collection time over traditional mark-sweep for 49% while keeping the heap size approximately the same.

Down for the count? Getting reference counting back in the ring

Reference counting and tracing are the two fundamental approaches that have underpinned garbage collection since 1960. However, despite some compelling advantages, reference counting is almost completely ignored in implementations of high performance systems today. In this paper we take a detailed look at reference counting to understand its behavior and to improve its performance. We identify key design choices for reference counting and analyze how the behavior of a wide range of benchmarks might affect design decisions. As far as we are aware, this is the first such quantitative study of reference counting. We use insights gleaned from this analysis to introduce a number of optimizations that significantly improve the performance of reference counting. We find that an existing modern implementation of reference counting has an average 30% overhead compared to tracing, and that in combination, our optimizations are able to completely eliminate that overhead. This brings the performance of reference counting on par with that of a well tuned mark-sweep collector. We keep our in-depth analysis of reference counting as general as possible so that it may be useful to other garbage collector implementers. Our finding that reference counting can be made directly competitive with well tuned mark-sweep should shake the community's prejudices about reference counting and perhaps open new opportunities for exploiting reference counting's strengths, such as localization and immediacy of reclamation.

Improved replication-based incremental garbage collection for embedded systems

We have developed an incremental compacting garbage collector for embedded Java systems. The collector divides the heap into equal sized pages and uses the segregated free lists for fast allocation. Collectors that have such a heap layout have a problem of fragmentation in allocating objects larger than the page size. We solve this problem by using the replication-based incremental compaction. The compactor evacuates all objects in one area, the evacuation area, of the heap, thereby creating a large chunk of free space. We developed an algorithm for choosing the evacuation area that effectively cures fragmentation. The compactor does not use any read-barriers. Instead, it uses a technique similar to the replication-based incremental copying collection. This needs forwarding pointers for all evacuated objects. Rather than introducing an extra field for each object, we use a hash table to store forwarding pointers. Evaluation of this garbage collector implemented in Sun's J2ME Java Virtual Machine showed that all the benchmarks used were able to run without memory starvation using the heap sizes of only 151%-286% of the maximum amount of live data plus 8 KB of the hash table. Experiments on a desktop computer, though it is not a platform for embedded systems, showed that the maximum pause time was shorter than 200 μs, which was comparable to that of our implementation of the snapshot-at-the-beginning collector without compaction. On an ARM processor, the runtime overhead was 1%-16% with 8.0% on average compared to the mark-sweep collector.

A parallel, incremental, mostly concurrent garbage collector for servers

Multithreaded applications with multigigabyte heaps running on modern servers provide new challenges for garbage collection (GC). The challenges for “server-oriented” GC include: ensuring short pause times on a multigigabyte heap while minimizing throughput penalty, good scaling on multiprocessor hardware, and keeping the number of expensive multicycle fence instructions required by weak ordering to a minimum.We designed and implemented a collector facing these demands building on the mostly concurrent garbage collector proposed by Boehm et al. [1991]. Our collector incorporates new ideas into the original collector. We make it parallel and incremental; we employ concurrent low-priority background GC threads to take advantage of processor idle time; we propose novel algorithmic improvements to the basic mostly concurrent algorithm improving its efficiency and shortening its pause times; and finally, we use advanced techniques, such as a low-overhead work packet mechanism to enable full parallelism among the incremental and concurrent collecting threads and ensure load balancing.We compared the new collector to the mature, well-optimized, parallel, stop-the-world mark-sweep collector already in the IBM JVM. When allowed to run aggressively, using 72% of the CPU utilization during a short concurrent phase, our collector prototype reduces the maximum pause time from 161 ms to 46 ms while only losing 11.5% throughput when running the SPECjbb2000 benchmark on a 600-MB heap on an 8-way PowerPC 1.1-GHz processors. When the collector is limited to a nonintrusive operation using only 29% of the CPU utilization, the maximum pause time obtained is 79 ms and the loss in throughput is 15.4%.

Mark-copy

Copying garbage collectors have a number of advantages over non-copying collectors, including cheap allocation and avoiding fragmentation. However, in order to provide completeness (the guarantee to reclaim each garbage object eventually), standard copying collectors require space equal to twice the size of the maximum live data for a program. We present a mark-copy collection algorithm (MC) that extends generational copying collection and significantly reduces the heap space required to run a program. MC reduces space overhead by 75--85% compared with standard copying garbage collectors, increasing the range of applications that can use copying garbage collection. We show that when MC is given the same amount of space as a generational copying collector, it improves total execution time of Java benchmarks significantly in tight heaps, and by 5--10% in moderate size heaps. We also compare the performance of MC with a (non-generational) mark-sweep collector and a hybrid copying/mark-sweep generational collector. We find that MC can run in heaps comparable in size to the minimum heap space required by mark-sweep. We also find that for most benchmarks MC is significantly faster than mark-sweep in small and moderate size heaps. When compared with the hybrid collector, MC improves total execution time by about 5% for some benchmarks, partly by increasing the speed of execution of the application code.

A parallel, incremental and concurrent GC for servers

Multithreaded applications with multi-gigabyte heaps running on modern servers provide new challenges for garbage collection (GC). The challenges for "server-oriented" GC include: ensuring short pause times on a multi-gigabyte heap, while minimizing throughput penalty, good scaling on multiprocessor hardware, and keeping the number of expensive multi-cycle fence instructions required by weak ordering to a minimum. We designed and implemented a fully parallel, incremental, mostly concurrent collector, which employs several novel techniques to meet these challenges. First, it combines incremental GC to ensure short pause times with concurrent low-priority background GC threads to take advantage of processor idle time. Second, it employs a low-overhead work packet mechanism to enable full parallelism among the incremental and concurrent collecting threads and ensure load balancing. Third, it reduces memory fence instructions by using batching techniques: one fence for each block of small objects allocated, one fence for each group of objects marked, and no fence at all in the write barrier. When compared to the mature well-optimized parallel stop-the-world mark-sweep collector already in the IBM JVM, our collector prototype reduces the maximum pause time from 284 ms to 101 ms, and the average pause time from 266 ms to 66 ms while only losing 10% throughput when running the SPECjbb2000 benchmark on a 256 MB heap on a 4-way 550 MHz Pentium multiprocessor.

Comparing mostly-copying and mark-sweep conservative collection

Many high-level language compilers generate C code and then invoke a C compiler for code generation. To date, most, of these compilers link the resulting code against a conservative mark-sweep garbage collector in order to reclaim unused memory. We introduce a new collector, MCC, based on an extension of mostly-copying collection .We analyze the various design decisions made in MCC and provide a performance comparison to the most widely used conservative mark-sweep collector (the Boehm-Demers-Weiser collector). Our results show that a good mostly-copying collector can outperform a mature highly-optimized mark-sweep collector when physical memory is large relative to the live data. A surprising result of our analysis is that cache behavior can have a greater impact on overall performance than either collector time or allocation time.