Dynamic object sampling for pretenuring

Abstract

Many state-of-the-art garbage collectors are generational, collecting the young nursery objects more frequently than old objects. These collectors perform well because young objects tend to die at a higher rate than old ones. However, these collectors do not examine object lifetimes with respect to any particular program or allocation site. This paper introduces low-cost object sampling to dynamically determine lifetimes. The sampler marks an object and records its allocation site every n bytes of allocation. The collector then computes per-site nursery survival rates. Sampling degrades total performance by only 3% on average for sample rates of 256 bytes in Jikes RVM, a rate at which overall lifetime accuracy compares well with sampling every object.An adaptive collector can use this information to tune itself. For example, pretenuring decreases nursery collection work by allocating new, but long-lived, objects directly into the mature space. We introduce a dynamic pretenuring mechanism that detects long-lived allocation sites and pretenures them, given sufficient samples. To react to phase changes, it occasionally backsamples. As with previous online pretenuring, consistent performance improvements on SPECjvm98 benchmarks are difficult to attain since only two combine sufficient allocation load with high nursery survival. Our pretenuring system consistently improves one of these, 213 javac, by 2% to 9% of total time by decreasing collection time by over a factor of two. Sampling and pretenuring overheads slow down all the others. This paper thus provides an efficient sampling mechanism that accurately predicts lifetimes, but leaves open optimization policies that can exploit this information.

Understanding the Impact of Collection Contracts on Design

“…[26][27][28][29]). Hirzel et al studied a suite of benchmarks and concluded that object connectivity correlates strongly with object lifetime [30].…”

Adaptive sampling-based profiling techniques for optimizing the distributed JVM runtime

“…In [27], sampling is used to characterize object allocation behavior while we use it to track and estimate sharing profile. In our system, space constraint is much tighter since we cannot store or transfer too much sampling result.…”

“…Thus, we have to start with a wide sampling gap (actually 4KB) and dynamically adjust it only when accuracy is not enough. In contrast, a much smaller sampling gap (256B) is chosen and fixed in [27].…”

A Phase-Adaptive Garbage Collector Using Dynamic Heap Partitioning and Opportunistic Collection

“…If an object escapes into the sink, the object is considered long-lived. Then, it is possible to avoid the processing of the long-lived object by copying it out of the frequently collected region similarly to the pretenuring studies [5], [7].…”

“…However, several studies [1]- [5] show that accurately predicting object lifetimes is difficult to achieve. The prediction mechanisms in those works either fail to attain performance improvement or are not applicable to various applications.…”