Overview of excelsior JET, a high performance alternative to java virtual machines

  • Authors:
  • V. Mikheev;N. Lipsky;D. Gurchenkov;P. Pavlov;V. Sukharev;A. Markov;S. Kuksenko;S. Fedoseev;D. Leskov;A. Yeryomin

  • Affiliations:
  • Excelsior, LCC, A.P. Ershov Institute of Informatics Systems, Russia;Excelsior, LCC, A.P. Ershov Institute of Informatics Systems, Russia;Excelsior, LCC, A.P. Ershov Institute of Informatics Systems, Russia;Excelsior, LCC, A.P. Ershov Institute of Informatics Systems, Russia;Excelsior, LCC, A.P. Ershov Institute of Informatics Systems, Russia;Excelsior, LCC, A.P. Ershov Institute of Informatics Systems, Russia;Excelsior, LCC, A.P. Ershov Institute of Informatics Systems, Russia;Excelsior, LCC, A.P. Ershov Institute of Informatics Systems, Russia;Excelsior, LCC, A.P. Ershov Institute of Informatics Systems, Russia;Excelsior, LCC, A.P. Ershov Institute of Informatics Systems, Russia

  • Venue:
  • WOSP '02 Proceedings of the 3rd international workshop on Software and performance
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper describes the architecture of Excelsior JET, a Java optimizing environment based upon ahead-of-time compilation of Java bytecode to platform-specific (native) code. Advantages of static compilation originate from the possibility of employing powerful resource-expensive optimizations resulting in much better code quality than in the case of dynamic (just-in-time) complation implemented in contemporary Java Virtual Machines (JVMs). Moreover, a noticeable innovation of JET is the Mixed Compilation Model that allows a compiled program to load classes dynamically as the Java Language Specification requires. Along with a highly optimizing compiler, JET includes a scalable run-time system that provides effective memory management with garbage collection, threading, etc. The experimental results presented in the paper were obtained on a variety of benchmarks and real-world Java applications. Tested against two most popular JVMs (Sun HotSpot and IBM JIT), programs optimized by JET demonstrated better performance and lower resource demand in most cases.