Title: Reducing Trace Selection Footprint for Large-scale Java Applications without Performance Loss

Authors: Peng Wu, Hiroshighe Hayashizaki, Hiroshi Inoue, Toshio Nakatani
Speaker: Peng Wu

Abstract:

When optimizing large-scale applications, striking the balance between
maximizing performance and preserving resources such as code size and
compilation time has always been a grand challenge. While recent
advances in trace compilation have demonstrated the feasibility of
building a Java trace JIT that can support large-scale applications
and have significantly reduced the performance gap to a mature method
JIT, the size control aspect of a trace compilation system remains
largely overlooked. For instance, we observe that about 40% of traces
formed for the DataCop 9.12 and DayTrader 2.0 benchmarks are
short-lived (e.g., with less than 500 execution count in steady-state
runs) and code duplication across traces are excessive (up to 20
times).

In this paper, we systematically study the size control problem for a
class of commonly used trace selection algorithms and propose a set of
techniques to reduce the footprint of trace selection without
incurring any performance loss. The crux of our approach is to target
redundancies in trace selection in the form of either short-lived
traces or unnecessary trace duplication. Using one of the best
performing selection algorithms as the baseline, we demonstrate that,
on the DaCapo 9.12 benchmarks and DayTrader 2.0 on WebSphere
Application Server 7.0, our techniques reduce the code size and
compilation time by 69% and the start-up time by 43% while retaining
the steady-state performance. On DayTrader 2.0, an example of
large-scale application, our techniques also improve the steady-state
performance by 10%.