Profile-based pretenuring

Abstract

Pretenuring can reduce copying costs in garbage collectors by allocating long-lived objects into regions that the garbage collector will rarely, if ever, collect. We extend previous work on pretenuring as follows: (1) We produce pretenuring advice that is neutral with respect to the garbage collector algorithm and configuration. We thus can and do combine advice from different applications. We find for our benchmarks that predictions using object lifetimes at each allocation site in Java programs are accurate, which simplifies the pretenuring implementation.(2) We gather and apply advice to both applications and Jikes RVM , a compiler and runtime system for Java written in Java. Our results demonstrate that building combined advice into Jikes RVM from different application executions improves performance, regardless of the application Jikes RVM is compiling and executing. This build-time advice thus gives user applications some benefits of pretenuring, without any application profiling. No previous work uses profile feedback to pretenure in the runtime system. (3) We find that application-only advice also consistently improves performance, but that the combination of build-time and application-specific advice is almost always noticeably better.(4) Our same advice improves the performance of generational, Older First, and Beltway collectors, illustrating that it is collector neutral. (5) We include an immortal allocation space in addition 2 • S. M. Blackburn et al.to a nursery and older generation, and show that pretenuring to immortal space has substantial benefit.

Decrypting the Java gene pool

“…Identification of suitable candidates may be through profiling programs [11,9,7], dynamic sampling at run-time [16,21], or program analysis [15].…”

“…Blackburn et al [9,7] extend this approach. They remove implementation dependency by normalising object lifetime as a multiple of the maximum volume of objects live at any time, live size.…”

“…Our scheme, on the other hand, leverages programspecific knowledge. Object lifetimes are recorded at a 64KB granularity in [9], as we do, but precisely in [7] using Merlin [17]. For combining advice in their algorithm, Blackburn et al cite the use of precise ages, plus ranking the importance of allocation sites by the volume they allocate rather than the product of the objects' volumes and ages, as the reason for the better consistency of results [7].…”

“…raytrace). Furthermore, there is evidence that, in Jikes RVM at least, a large fraction of long-lived objects are allocated at the start of the program [7], for which sampling-based pretenuring would be too late.…”

“…Blackburn et al [9,7] use object birth and death statistics, gathered from instrumented programs in an off-line phase, to advise the allocator as to whether particular allocation sites (points in the program which allocate new objects) are expected to create objects with short, long or immortal lifetimes. Acquiring advice for self-prediction [4] of particular programs is extremely expensive.…”