Software speculative multithreading for Java

  • Authors:
  • Christopher J. F. Pickett

  • Affiliations:
  • McGill University, Montreal, PQ, Canada

  • Venue:
  • Companion to the 22nd ACM SIGPLAN conference on Object-oriented programming systems and applications companion
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

We apply speculative multithreading to sequential Java programs in software to achieve speedup on existing multiprocessors. A common speculation library supports both Java bytecode interpreter and JIT compiler implementations. Initial profiling results indicate three main optimizations: adaptive return value prediction, online fork heuristics, and in-order nested method level speculation.