Pipelined Java Virtual Machine Interpreters

Abstract

Abstract. The performance of a Java Virtual Machine (JVM) interpreter running on a very long instruction word (VLIW) processor can be improved by means of pipelining. While one bytecode is in its execute stage, the next bytecode is in its decode stage, and the next bytecode is in its fetch stage. The paper describes how we implemented threading and pipelining by rewriting the source code of the interpreter and several modifications in the compiler. Experiments for evaluating the effectiveness of pipelining are described. Pipelining improves the execution speed of a threaded interpreter by 19.4% in terms of instruction count and 14.4% in terms of cycle count. Most of the simple bytecodes, like additions and multiplications, execute in four cycles. This number corresponds to the branch latency of our target VLIW processor. Thus most of the code of the interpreter is executed in branch delay slots.

Optimization strategies for a java virtual machine interpreter on the cell broadband engine

“…The most closely related work to ours is from Hoogerbrugge and Augusteijn who developed a Java interpreter for the Philips Trimedia VLIW processor [6]. This processor uses delayed branches, which are analogous to the hinted branches found in the Cell BE SPE architecture.…”

“…The Philips TriMedia VLIW Media Processor has some analogous features to the Cell BE SPE. Hoogerbrugge et al [6] present a pipeline interpter design in their paper for the Trimedia Processor. We have implemented their techniques and improved on the concept by adding operand fetching to the pipeline (see Section 6.1).…”

“…Hoogerbrugge et al [6] presented a pipelined interpreter for the TriMedia VLIW Media Processor. They showed that their techniques could decrease VLIW instruction count by up to 19.4% and cycle count by up to 14.4%.…”

“…Section 6.1 investigates a technique first proposed by Hoogerbrugge et al [6] called interpreter pipelining. Section 6.2 presents some unique techniques for the dispatch of JVM instructions.…”

Performance Research and Optimization on CPython’s Interpreter

“…J Hoogerbrugge and L Augusteijn [17], [18] have proposed that software pipelining interpreters is a way to reduce dispatch branch cost on architectures with split indirect branches. In addition, Subroutine threading [19] has also been proposed to avoid the overheads of indirect branches in intrepreter implementations.…”